JP2001157979A - Robot device, and control method thereof - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a robot device, and a control method capable of improving entertainment performance. SOLUTION: In this robot device to determine an output action based on an action model comprising a probability condition transition model, a condition space generating means is provided to generate a new condition by determining various setting items respectively at random, and generate a new condition space comprising the condition. In this control method for a robot device to determine an output action based on an action model comprising a probability condition transition model, a first step to generate a new condition by determining various setting items respectively at random, and generate a new condition space comprising the condition, and a second step to let the robot device to take an action in accordance with the generated new condition space are provided.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a robot apparatus and a control method therefor, and is suitably applied to, for example, a pet robot.

[0002]

2. Description of the Related Art In recent years, a pet robot of a four-legged walking type has been proposed and developed by the present applicant. Such a pet robot has a shape similar to a dog or cat bred in a general household, and can act autonomously in response to a user's action such as "hitting" or "patting" or the surrounding environment. It was done in.

[0003]

However, in such a pet robot, although it behaves autonomously as described above, the behavior pattern that is ultimately expressed is selected from the behavior patterns created by the developer. Therefore, the pet robot does not generate a completely new behavior pattern.

[0004] For this reason, in such a pet robot, the behavior pattern output over a long period of time becomes a repetition of the same behavior pattern, and there is a possibility that the user may feel bored.

As one method for solving such a problem, a method of increasing the number of behavior patterns expressed by a pet robot is conceivable. However, according to this method, the more the number of behavior patterns, the longer the time required to create a behavior model. In addition, there is a problem that requires a large-capacity memory for storing the behavior pattern.

[0006] Therefore, if the above-mentioned pet robot can be provided with, for example, a function of self-generating a behavior pattern, it is considered that such a problem can be solved at once and the entertainment property can be further improved.

The present invention has been made in view of the above points, and an object of the present invention is to propose a robot apparatus capable of improving entertainment characteristics and a control method thereof.

[0008]

According to the present invention, there is provided a robot apparatus for determining an output action based on an action model which is a stochastic state transition model, wherein various setting items are determined at random. And a state space generating means for generating a new state space consisting of the state.
As a result, this robot device can self-generate a new behavior pattern with the generation of a new state space.

According to the present invention, in a control method of a robot apparatus for determining an output action based on an action model comprising a stochastic state transition model, a new state is generated by randomly determining various setting items. The first step of generating a new state space including the state and the second step of causing the robot device to exhibit an action according to the generated new state space are provided. As a result, according to the control method of the robot device, a new behavior pattern can be generated by the generation of a new state space.

[0010]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below in detail with reference to the drawings.

(1) First Embodiment (1-1) Configuration of Pet Robot According to First Embodiment In FIG. 1, reference numeral 1 denotes a pet robot according to the first embodiment as a whole, The leg units 3A to 3D are connected to the front, rear, left and right of the unit 2, respectively.
The head unit 4 and the 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 Only Memory) 12, PC (Personal Com
puter) A control unit 16 formed by connecting a card interface circuit 13 and a signal processing circuit 14 to each other via an internal bus 15 and a battery 17 as a power source of the pet robot 1 are housed therein. . The body unit 2 has an angular velocity sensor 18 for detecting the direction and the acceleration of the movement of the pet robot 1.
And an acceleration sensor 19 are also stored.

The head unit 4 also detects a charge coupled device (CCD) camera 20 for capturing an image of an external situation, and a pressure received by a physical action such as "stroke" or "hit" from the user. A touch sensor 21, a distance sensor 22 for measuring a distance to an object located ahead, a microphone 23 for collecting external sound, and a speaker 24 for outputting a sound such as a cry. , An LED (Light Emitting Diode) (not shown) corresponding to the “eyes” of the pet robot 1 and the like 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
Each coupling portion of the head unit 4 and body unit connecting portion 2, and the tail unit actuator 25 _1, respectively, etc. The consolidated portion of the tail 5A of freedom minutes 5
To 25 _n and potentiometers 26 _{1 to} 26 _n are provided.

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

At this time, 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 receives these sequentially into the internal bus 15.
, And sequentially stored in a predetermined position in the DRAM 11. 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 used when the CPU 10 controls the operation of the pet robot 1 thereafter.

In practice, when the power of the pet robot 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 the PC card interface. The data is read out via the circuit 13 or directly and stored in the DRAM 11.

Further, the CPU 10 thereafter, based on the sensor data, image data, audio data, and remaining battery power data sequentially stored in the DRAM 11 from the signal processing circuit 14 as described above, and from the user and the surroundings. Judge whether or not there is an instruction for and action.

Further, the CPU 10 determines the result of this determination and D
And it determines a subsequent action based on the control program stored in the RAM 11, by driving the actuator 25 ₁ to 25 _n required based on the determination result,
The head unit 4 can be swung up and down, left and right, the tail 5A of the tail unit 5 can be moved, and each of the leg units 3A to 3A.
D is driven to perform an action such as walking.

At this time, the CPU 10 generates audio data as required 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.

As described above, the pet robot 1 is capable of acting autonomously in accordance with its own and surrounding conditions, instructions and actions from the user.

(1-2) Software Configuration of Control Program The software configuration of the above-described control program in the pet robot 1 is shown in FIG. In FIG.
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 a CCD camera 2
0 (FIG. 2), an object such as a timer, which is allowed to directly access hardware used in a normal computer, and performs processing in response to an interrupt from the corresponding hardware.

The robotic server object 32 is located above the device driver layer 30 and includes, for example, the various sensors and actuators 25 ₁ described above.
A virtual robot 33 comprising a software group that provides an interface for accessing the hardware, such as to 25 _n, the power manager 34 made of a software suite for managing the power supply switching, the other various device drivers Device driver manager 35, which is a group of software to be managed, and pet robot 1
And a designed robot 36 which is a software group for managing the mechanism.

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 pet robot 1 such as image processing and sound processing. I have.
Further, the application layer 41 is located in a layer above the middleware layer 40, and
It is composed of a software group for determining the behavior of the pet robot 1 based on the processing result processed by each software group constituting the layer 40.

FIGS. 4 and 5 show specific software configurations of the middleware layer 40 and the application layer 41, respectively.

In the middleware layer 40,
As is clear from FIG. 4, for scale recognition, for distance detection,
Recognition system 5 including signal processing modules 50 to 55 for posture detection, touch sensor, motion detection, and color recognition, input semantics converter module 56, and the like.
7 and the output semantics converter module 57 and for attitude management, tracking, motion playback,
An output system 65 includes signal processing modules 58 to 64 for walking, falling back, turning on LEDs, and reproducing sounds.

In this case, each of the signal processing modules 50 to 55 of the recognition system 57 is provided by the virtual robot 33 of the robotic server object 32 to the DRAM 11.
The corresponding data among the sensor data, image data, and audio data read out from FIG. 2 (FIG. 2) is fetched, predetermined processing is performed based on the data, and the processing result is provided to the input semantics converter module 56.

The input semantics converter module 56, based on the processing results given from each of the signal processing modules 50 to 55, "detects a ball"
"Detected fall,""stroke,""beaten,"
"I heard the domeso scale,""Detected a moving object."
Alternatively, it recognizes the situation of itself and surroundings, such as "detected an obstacle", and commands and actions from the user, and outputs the recognition result to the application layer 41 (FIG. 2).

As shown in FIG. 5, the application layer 41 includes five modules: a behavior model library 70, a behavior switching module 71, a learning module 72, an emotion model 73, and an instinct model 74.

In this case, the behavior model library 70 includes
As shown in FIG. 6, "when the battery level is low", "when the vehicle returns to the fall", "when avoiding obstacles", "when expressing emotion", "when the ball is detected", and the like. Respectively correspond to several pre-selected condition items, and independent behavior models 70 _{1 to} 7 respectively.
0 _n is provided.

These behavior models 70 ₁ to 70
_n is stored in the emotion model 73 as described later as necessary when a recognition result is given from the input semantics converter module 56 or when a certain time has elapsed since the last recognition result was given. The next action is determined with reference to the corresponding emotion parameter value and the corresponding desire parameter value held in the instinct model 74, and the determination result is output to the action switching module 71.

In this embodiment, each of the behavior models 70 ₁ to 70 _n uses one node (state) NODE _{0 to} NO as shown in FIG.
Based whether to transition from DE _n to any other node NODE ₀ ~NODE _n the transition probability P ₁ to P _n which is set respectively arc ARC ₁ ~ARC _n1 to connect each node NODE ₀ ~NODE _n An algorithm called a probabilistic automaton that determines stochastically is used.

Specifically, each of the behavior models 70 ₁ to 70
_n corresponds to each of the nodes NODE _{0 to} NODE _n forming their own behavior models 70 ₁ to 70 _n , respectively, and FIG. 8 for each of these nodes NODE _{0 to} NODE _n
Has a state transition table 80 as shown in FIG.

In the state transition table 80, the node NO
Input events (recognition results) as transition conditions in DE _{0 to} NODE _n are listed in order of priority in the row of “input event name”, and further conditions for the transition conditions are described in the rows of “data name” and “data range”. It is described in the corresponding column.

Therefore, in the node NODE ₁₀₀ represented by the state transition table 80 of FIG. 8, "ball detected (BALL)"
When the recognition result is given, the "size (SIZE)" of the ball given together with the recognition result is in the range of "0 to 1000", or the recognition of "obstacle detected (OBSTACE)" is given. When the result is given, the condition for transition to another node is that the "distance" to the obstacle given together with the recognition result is in the range of "0 to 100". .

In the node NODE ₁₀₀ , even when the recognition result is not input, the behavior models 70 _{1 to} 70
0 _n of the parameter values of each emotion and each desire held in the emotion model 73 and the instinct model 74 that are periodically referred to, the “joy (JO)” held in the emotion model 73.
When the parameter value of any of "Y)", "Surprise", or "Sadness" is in the range of "50 to 100", it is possible to make a transition to another node.

Further in the state transition table 80, with the node name that can transition from the node NODE ₀ ~NODE _n in the column of "transition destination node" in the column of "probability of transition to another node" is listed, the "input Other nodes NODE _{0 to} N that can transition when all the conditions described in the rows of “event name”, “data value”, and “data range” are met
The transition probability to ODE _n is described in a corresponding part in the column of “transition probability to another node”, and the node N
The action to be output when transitioning from ODE _{0 to} NODE _n is “output action” in the column of “transition probability to another node”.
Is described in the line. Note that the sum of the probabilities of the respective rows in the column of “transition probability to another node” is 100 [%].

Therefore, in the node NODE ₁₀₀ represented by the state transition table 80 in FIG.
L) "and the" SIZE "of the ball is" 0 ".
Is given in the range of “node NODE” with a probability of “30 [%]”.
₁₂₀ (node 120) ”, and then“ ACTIO ”
The action of "N1" will be output.

Each of the behavior models 70 ₁ to 70 _n is configured such that a number of nodes NODE _{0 to} NODE _n described as such a state transition table 80 are connected, and the input semantics converter module 56 For example, when a recognition result is given, the next action is stochastically determined using the state transition table 80 of the corresponding nodes NODE ₀ to NODE _n , and the determination result is output to the action switching module 71. ing.

The action switching module 71, among the actions respectively output from the behavior model 70 ₁ to 70 _n of the action model library 70, output from the predetermined high priority behavior model 70 ₁ to 70 _n A command for selecting an action and executing the action (hereinafter, referred to as an action command) is transmitted to the middleware layer 40.
To the output semantics converter 57. Incidentally in this embodiment, is denoted behavioral models 70 ₁ more to 70 _n priority lower is set higher in FIG.

Further, the action switching module 71, based on the action completion information provided from the output semantics converter 57 after the action is completed, notifies the learning module 72, the emotion model 73, and the instinct model 7 that the action has been completed.
Notify 4.

On the other hand, the learning module 72 inputs the recognition result of the teaching received from the user, such as “hit” or “stroke”, among the recognition results given from the input semantics converter 56.

Then, based on the recognition result and the notification from the action switching module 71, the learning module 72 lowers the probability of occurrence of the action when "beaten (scolded)" and "strokes (praise)". was) "sometimes to increase the expression probability of that action, behavior model 70 _1-7 corresponding in behavioral model library 70
Change the corresponding transition probability of 0 _n .

On the other hand, the emotion model 73 indicates "joy (joy)
) "," Sadness "," anger "
) "," Surprise "," disgust "
For a total of six emotions of “fear” and “fear”, a parameter indicating the strength of the emotion is stored for each emotion. Then, the emotion model 73 converts the parameter values of each of these emotions from the specific recognition results such as “hit” and “stroke” given from the input semantics converter module 56 and the elapsed time and action switching module 71, respectively. The notification is sequentially updated based on the notification and the like.

Specifically, the emotion model 73 includes a recognition result from the input semantics converter 56 and a degree (preset) at which the behavior of the pet robot 1 acts on the emotion at that time. The held parameter values of each desire and the degree at which the behavior of the pet robot 1 acts on the emotion (preset), the degree of suppression and stimulation received from other emotions, the elapsed time ΔE [t], the current parameter value of the emotion E [t], and the rate of change of the emotion according to the recognition result, etc. hereinafter, the coefficient representing the same is referred to as sensitivity) as k _e, the following expression in a predetermined cycle

[0048]

(Equation 1)

Is used to calculate the parameter value E [t + 1] of the emotion in the next cycle.

The emotion model 73 updates the parameter value of the emotion by replacing the calculation result with the parameter value E [t] of the emotion. It is to be noted that the parameter value of the emotion to be updated in response to each recognition result or the notification from the action switching module 71 is determined in advance. For example, when a recognition result such as “hit” is given, “anger” is given. The emotion parameter value has increased,
When a recognition result such as “stroke” is given, the parameter value of the emotion of “joy” increases.

On the other hand, the instinct model 74 includes “exercise (exersise)”, “affection (afection)”, and “appetite (ap
For each of the four independent desires "petite" and "curiocity", a parameter indicating the strength of the desire is held for each of the desires. Then, the instinct model 74 converts the parameter values of the desire into the recognition result given from the input semantics converter module 56, the elapsed time and the action switching module 71, respectively.
The information is sequentially updated based on a notification or the like.

More specifically, the instinct model 74 is “exercise desire”,
Regarding “affection desire” and “curiosity”, the change amount of the desire calculated by a predetermined arithmetic expression based on the action output, elapsed time, and recognition result of the pet robot 1 is ΔI
[K], I [k] of the current parameter value of the desire, the coefficient representing the sensitivity of the desire as k _i, the following expression in a predetermined cycle

[0053]

(Equation 2)

Is used to calculate the parameter value I [k + 1] of the desire in the next cycle, and replaces the result of this operation with the current parameter value I [k] of the desire to update the parameter value of the desire. I do. It is to be noted that the parameter value of the desire to be changed with respect to the behavior output, the recognition result, and the like is determined in advance. Parameter value decreases.

In the instinct model 74, the “appetite” is calculated based on the battery remaining amount data provided through the input semantics converter module 56 as the battery remaining amount as B _L at a predetermined cycle.

[0056]

(Equation 3)

The parameter value I [k] of "appetite" is calculated by the
The parameter value of the “appetite” is updated by replacing with [k].

In the present embodiment, the parameter values of each emotion and each desire are regulated so as to fluctuate in the range of 0 to 100, and the coefficients k _e and k _i
Is also set individually for each emotion and each desire.

On the other hand, as shown in FIG. 4, the output semantics converter module 57 of the middleware layer 40 is "forward" and "pleased" given from the action switching module 71 of the application layer 41 as described above. , "Sound" or "tracking (follow the ball)" to the corresponding signal processing modules 58 to 64 of the output system 65.

The signal processing modules 58 to 6
4, given the behavior command based on the action command, and servo command value to be applied to the actuator 25 ₁ to 25 _n (Fig. 2) corresponding to perform that action, the output from the speaker 24 (FIG. 2) Generating sound data of the sound to be played and / or driving data to be given to the LED of the “eye”,
These data are transferred to the virtual robot 33 of the robotic server object 32 and the signal processing circuit 14.
The corresponding actuators 25 ₁ to 25 ₁ through (FIG. 2)
25 _n , sequentially to the speaker 24 or the LED.

As described above, the pet robot 1 can perform an autonomous action in accordance with the self and surrounding conditions, and instructions and actions from the user, based on the control program. .

(1-3) Behavior Pattern Self-Generation Function In addition to the above configuration, in the case of the pet robot 1, the behavior pattern (a series of behaviors) is self-generated and the generated behavior pattern is matched to the user's preference. It is designed to be able to change.

That is, in the pet robot 1, each of the behavior models 70 ₁ to 70 ₁ to
70 out of _n, the specific behavior model, such as a ball corresponding behavior model 70 _n and emotion behavior model _n-1 (hereinafter, referred to as specific behavior model 70 k ₍₁₎₎ In the so shown in FIG. 9 In addition, a plurality of nodes (hereinafter, referred to as starting nodes) NODE _{st serving} as a starting point of a self-generated action pattern are provided.

When the specific behavior model 70 _{k (1)} transits to the origin node NODE _st , at a certain probability, a random number of new nodes (hereinafter, referred to as newly generated nodes) NODE _{new ( 1) to} NODE _{new (3)} , and these newly generated nodes NODE _{new (1) to} N
A node sequence 90 for sequentially determining output actions as shown in FIG. 10 composed of ODE _{new (3)} is generated.

In this case, each newly created node NODE
_{The generation of new (1) to} NODE _{new (3)} is performed by newly generating the state transition table 80 described above with reference to FIG. At this time, node names, transition conditions (items described in the lines of “input event name”, “data name”, and “data range”), transition destinations, priority of transition conditions, output actions, transition probabilities, etc. Various setting items are randomly selected and determined using a random number or the like.

At this time, the newly created node NODE
_{From new (1) to} NODE _{new (3)} , the next newly created node NOD
A plurality of types (four types in this embodiment) of actions are arbitrarily selected as output actions when transitioning to E _{new (2)} and NODE _{new (3)} , and the transition probabilities are respectively selected for these actions. Is assigned. Moreover always includes origin node NODE _st is the transition destination of each new generation node _{NODE new (1) ~NODE new (} 3), the origin node NODE _st
The transition probability and the output action are also assigned to the transition to.

[0067] Furthermore the origin node NODE _st this time, along with the first new generation node NODE _{new (2)} is added at node column 90 as a transition destination, when transitioning to the new generation node NODE _{new (2)} Four types of actions are selected as output actions, and a transition probability is randomly assigned to each of these actions. For this reason, for the origin node NODE _st , the transition probabilities at necessary locations are changed to appropriate values so that the sum of the transition probabilities of each row of the state transition table 80 becomes 100 [%].

Then, the special behavior model 70 _{k (1)} follows the origin node N according to the node sequence 90 thus generated.
From ODE _st to new generation node NODE _{new (1),} also newly generated node _{NODE new (1) ~NODE new (} 3) from the next newly generated node _{NODE new (2), NODE new} (3) or the source node NODE _st Then, the output behavior of the pet robot 1 is determined stochastically sequentially by sequentially transitioning the nodes, and the determination result is sequentially output to the behavior switching module 71 (FIG. 5).

As a result, a newly generated action pattern is expressed by the pet robot 1 based on the output of the special action model 70 _{k (1)} . Thus, the pet robot 1 generates and expresses a new behavior pattern.

Further, in the pet robot 1, when the behavior pattern based on the node sequence 90 generated as described above is expressed, the user “strokes” or “hits”.
(Hereinafter, referred to as teaching), the transition probability corresponding to some actions immediately before that increases or decreases by the action of the learning module 70 (FIG. 5) as described above.

As a result, in the pet robot 1, while repeating the behavior pattern based on the above-described node sequence 90, the behavior whose appearance probability has been increased by the user being “patched (praised)” by each user appears. The behavior pattern is likely to change to a pattern in which these behaviors are combined.

However, in this state, the starting node NODE
When transitioning from _st to the next newly created node NODE _{new (1)} or from the newly created nodes NODE _{new (1) to} NODE _{new (3)} , the next newly created nodes NODE _{new (2)} and NODE _{new (3)}
Alternatively, when transiting to the origin node NODE _st , only one of the four types of behaviors selected at the initial stage is expressed.

Therefore, in each special behavior model 70 _{k (1)} , each newly created node NODE _{new (1) to} NODE
_With regard to _{new (3)} , an action having a low transition probability in the state transition table 80 is replaced with a new action at a predetermined cycle, and the transition probability for each action is set again randomly.

Thus, in the pet robot 1, such an action exchange and a new node sequence 90 are performed.
While repeating the expression of the action pattern based on, the action pattern to be expressed changes to the pattern preferred by the user according to the instruction given by the user.

As described above, in the pet robot 1, the behavior pattern can be generated by itself, and the generated behavior pattern can be changed according to the user's preference.

(1-4) Operation and Effect of the Present Embodiment In the above configuration, the pet robot 1 generates a node sequence 90 including a plurality of newly generated nodes NODE _{new (1) to} NODE _{new (3).} At the same time, the output behavior is stochastically and sequentially determined according to the node sequence 90, and the behavior pattern is expressed based on the determination result.

At this time, the pet robot 1 increases or decreases the occurrence probability of the output action based on the instruction given by the user at this time, replaces the action with a low occurrence probability with a new action, and redefines the occurrence probability for each action. Thus, the action pattern generated based on the node sequence 90 is sequentially changed.

Therefore, the pet robot 1 can generate a completely new behavior pattern by itself and change this behavior pattern sequentially according to the user's preference, so that the behavior pattern can be prevented from being monotonous.

Further, in the pet robot 1, since the action patterns generated in this manner are sequentially changed according to the user's preference, it is possible to provide the user with the fun of “fostering”.

According to the above configuration, a node sequence 9 including a plurality of newly created nodes NODE _{new (1) to} NODE _{new (3)}
0 and stochastically determine output actions sequentially in accordance with the node sequence 90 and express a series of actions based on the determination result, thereby avoiding monotonicity of the action pattern and thus amusement performance. Can be realized.

(2) Second Embodiment (2-1) Action Pattern Self-Generation Function According to Second Embodiment In FIG. 1, reference numeral 100 denotes a pet robot according to the second embodiment as a whole, Except that the method of generating the behavior pattern in the behavior model 70 _{k (2)} is different,
It is configured similarly to the pet robot 1 according to the embodiment.

[0082] That is, in the case of this pet robot, a specific behavior model 70 _{k (2)} (Fig. 9) is the origin node NODE _st
When a specific recognition result accompanied by incidental information such as “detected an obstacle” or “detected a ball” is given in a state where the state has changed to the starting node NODE _st as shown in FIG. Is generated in the same manner as the node sequence 90 (FIG. 10) of the first embodiment described above, and each branch in the node sequence 101 is generated. A specific branch condition according to the recognition result is given to the location.

For example, in the recognition result such as “detected an obstacle”, information on “distance” to the obstacle and information on “direction” of the obstacle (output from a yaw direction potentiometer at the neck) are output. Angle information). Therefore, when a specific behavior model 70 _{k (2)} is given a recognition result such as “detected an obstacle,” FIG.
, The node sequence 101 having two branches for “distance” and “direction” is generated with the origin node NODE _st as a starting point, and the “distance” is randomly set for the “distance” branch point. A branch condition for determining whether or not the direction is larger than the set first threshold value S ₁ is given, and the “direction” (angle) of the “direction” branch point is larger than the appropriately set second threshold value S _2. Is given.

Then, the specific behavior model 70 _{k (2)} is thereafter changed from the origin node NODE _st to the newly created node NODE ₍₁₎ or the newly created node NODE ₍₂₎ according to the node sequence 101 thus created. Generation node NOD
E ₍₁₎ or newly created node NODE ₍₂₎ to newly created node NODE ₍₃₎ or newly created node NODE ₍₄₎
Or, to the newly created node NODE ₍₅₎ or the newly created node NODE ₍₆₎ , and further to the newly created node NODE ₍₃₎
~ Newly generated node NODE ₍₅₎ to start node NODE
The output behavior of the pet robot 100 is determined stochastically sequentially by sequentially transitioning the nodes, such as to _st , and the determination result is sequentially output to the behavior switching module 71 (FIG. 5).

As a result, a newly generated action pattern is expressed by the pet robot 100 based on the output of the special action model 70 _{k (2)} . In this manner, the pet robot 100 generates and expresses a new behavior pattern.

In the pet robot 100, when the user gives a teaching while expressing the action pattern based on the node sequence 101 generated as described above, as described above, the learning module 70 ( By the operation of FIG. 5), the respective transition probabilities corresponding to some previous actions increase or decrease. In addition, the learning module 70 also determines the thresholds (first and second thresholds S ₁ , S ₂ ) of some branch locations immediately before the node sequence 101 in the specific behavior model 70 _{k (2)} . Increase or decrease the value.

As a result, in the pet robot 100, while repeating the behavior pattern based on the above-described node sequence 101, the behavior whose probability of occurrence has been increased by the user being “stroked (praised)” by each user appears. And the first and second thresholds S ₁ and S ₂ are also changed by being “stroked (praised)” by the user.

However, in this case, the starting node NODE
When transitioning from _st to node NODE ₍₁₎ or node NODE ₍₂₎ , or from newly created node NODE ₍₁₎ or newly created node NODE _(2), to newly created node NODE ₍₃₎ or newly created node NODE ₍₄₎ Alternatively, initially selected when transitioning to the newly created node NODE ₍₅₎ or the newly created node NODE ₍₆₎ , and when transitioning from the newly created node NODE ₍₃₎ to the newly created node NODE ₍₅₎ to the start node NODE _st Only one of the actions performed is expressed.

Therefore, in each of the special behavior models 70 _{k (2)} , for each of the newly created nodes NODE _{(1) to} NODE ₍₆₎ , an action having a low transition probability in the state transition table 80 at a predetermined cycle is regarded as a new action. At the same time, the transition probability for each action is randomly set again.

Thus, in the pet robot 100, such an action exchange and a new node sequence 1 are performed.
While repeating a series of actions based on 01, the action pattern to be developed gradually becomes an action pattern when an optimal decision is made based on the accompanying information in accordance with the instruction given by the user.

In this manner, the pet robot 100
In, self-generating behavior patterns with judgment,
The behavior pattern can be changed based on the instruction given by the user.

(2-2) Operation and Effect of the Present Embodiment In the above configuration, when a specific recognition result is given, the specific behavior model 70 _{k (2)} determines the starting node NODE _st according to the recognition result. , A node sequence 101 having a tree structure having a number of layers corresponding to the number of incidental information items is generated, and a specific branch condition according to the recognition result is given to each branch position in the node sequence 10, thus obtained. An action pattern is generated based on the node sequence 101.

At this time, the pet robot 100 sequentially changes the occurrence probability of the output action and the branch condition set for each branch point in the node sequence 101 based on the teaching given by the user at this time. The action is replaced with a new action, and the action pattern generated based on the node sequence 101 is sequentially changed by redefining the occurrence probability for each action.

Therefore, the pet robot 100 generates a completely new behavior pattern by the processing of the specific behavior model 70 _{k (2)} , and optimizes the behavior pattern based on the accompanying information accompanying the recognition result. Since it is possible to sequentially change the behavior pattern when making an appropriate decision, in addition to avoiding the monotony of the behavior pattern, it is possible to generate an optimal behavior pattern according to the situation.

According to the above configuration, the node sequence 101 having a tree structure composed of a plurality of newly created nodes NODE _{new (1) to} NODE _{new (6)} is generated, and the node sequence 10 is generated.
1, a predetermined branch condition is given to each branch point, output actions are determined stochastically sequentially according to the node sequence 101, and a series of actions are expressed based on the determination result. It is possible to generate an optimal behavior pattern according to the situation in addition to avoiding the occurrence of a pet, and thus a pet robot that can improve amusement properties can be realized.

(3) Other Embodiments In the first and second embodiments described above, a case has been described in which the present invention is applied to a four-legged walking type pet robot. The present invention is not limited to this, and can be widely applied to various other types of robot devices.

In the first and second embodiments described above, a new state (a newly created node NODE _{new (1)} to
NODE _{new (6)} ), and a specific action model 7 as a state space generating means for generating a new state space (node trains 90 and 101) including the new state.
0 _{k (1)} and 70 _{k (2)} (and CPU 10) are shown in FIGS.
However, the present invention is not limited to this, and a new spatial state generated by the specific behavior models 70 _{k (1)} and 70 _{k (2)} is described. In addition, various other forms can be widely applied.

Further, in the above-described second embodiment, a case has been described in which the node sequence 101 as shown in FIG. 11 is generated. However, the present invention is not limited to this, and FIG.
As shown in FIG. 2, the node column 101 has another tree structure.
Node string 1 by linking two nodes 102
03 generates, may proceed to NODE _{new (6)} is a starting node of the subsequent node column 102 without returning to the starting point node NODE _st when it reaches the example node NODE _{new (5).} By doing so, the cycle of judgment and action can be performed twice, and a more complicated action pattern can be learned accordingly.

[0099]

As described above, according to the present invention, in a robot apparatus for determining an output action based on an action model comprising a stochastic state transition model, a new state is determined by randomly determining various setting items. Is generated, and a state space generating means for generating a new state space including the state is provided, whereby a new action pattern can be self-generated along with the generation of a new state space. A robot device capable of improving amusement can be realized.

Further, in the above-described embodiment, in the control method of the robot apparatus for determining the output action based on the action model which is the stochastic state transition model, various setting items are respectively determined at random and a new state is determined. And a first step of generating a new state space including the state, and a second step of causing the robot apparatus to exhibit an action corresponding to the generated new state space. Accordingly, a new behavior pattern can be generated by the robot device by itself, and the new behavior pattern can appear, thereby realizing a robot device capable of improving amusement properties.

[Brief description of the drawings]

FIG. 1 is a perspective view illustrating an external configuration of a pet robot according to first and second embodiments.

FIG. 2 is a block diagram illustrating a circuit configuration of the pet robot.

FIG. 3 is a conceptual diagram showing a software configuration of a control program.

FIG. 4 is a conceptual diagram showing a software configuration of a middleware layer.

FIG. 5 is a conceptual diagram showing a software configuration of an application layer.

FIG. 6 is a conceptual diagram serving to explain an action model library.

FIG. 7 is a schematic diagram illustrating a stochastic automaton.

FIG. 8 is a table showing a state transition table.

FIG. 9 is a conceptual diagram serving to explain the configuration of a specific behavior model.

FIG. 10 is a conceptual diagram explaining a node sequence generated in the first embodiment;

FIG. 11 is a conceptual diagram serving to explain a node sequence generated in the second embodiment.

FIG. 12 is a conceptual diagram for explaining another embodiment.

[Explanation of symbols]

1, 100 ... pet robot, 10 ... CPU, 16
…… Control unit 33 …… Virtual robot, 40
... Middleware layer, 41, 100 Applicationware, 70 Behavior model library,
70 ₁ to 70 _n ... action model, 70 _{k (1)} , 70 _{k (2)} ,
…… Specific action model, 90, 101-103 …… Node sequence, NODE _st …… Start point node, NODE _{new (1)} -NO
DE _{new (6)} …… New generation node.

 ────────────────────────────────────────────────── ─── Continued on the front page (72) Inventor Masahiro Fujita 6-35, Kita-Shinagawa, Shinagawa-ku, Tokyo F-term within Sony Corporation (reference) 3F059 AA00 BA02 BB06 CA05 DA05 DA09 DB02 DC01 DD01 DD11 FA03 FA05 FB01 FC02 FC03 FC14 FC15 3F060 AA00 CA14 GA05 GA13 GB06 GB25 GD12 GD14 GD15 5B049 BB07 DD00 DD03 EE03 EE07 EE41 FF00 FF06 FF07 FF08 5H004 GA26 GB16 HA07 HB07 HB08 HB09 HB15 HB15 JA03 JA05 JB06 MAB05 KD52 MA06 KD52

Claims (8)

[Claims] 1. A robot apparatus having an action model comprising a stochastic state transition model and determining an output action based on the action model, wherein a new state is generated by randomly determining various setting items. And a state space generating means for generating a new state space including the state. 2. The robot apparatus according to claim 1, wherein the setting items are a transition destination, a transition condition, a transition probability, and the output action. 3. The state space generating means changes a corresponding transition probability in the state space according to a predetermined external input.
The robot device according to item 1. 4. The state space has a tree structure in which a predetermined branch condition is given to each branch point, and the state space generating means determines the corresponding branch condition in response to the predetermined external input. The robot device according to claim 3, wherein the robot device is changed. 5. A control method for a robot apparatus having an action model comprising a stochastic state transition model and determining an output action based on the action model, wherein a new state is determined by randomly determining various setting items. A first step of generating and generating a new state space including the state, and a second step of causing the robot apparatus to express an action according to the generated new state space. A method for controlling a robot device. 6. The method according to claim 5, wherein the setting items are a transition destination, a transition condition, a transition probability, and the output action. 7. The method according to claim 5, further comprising a third step of changing the corresponding transition probability in the new state space according to a predetermined external input. . 8. The state space has a tree structure in which a predetermined branch condition is given to each branch point. In the third step, the corresponding branch condition is determined according to the predetermined external input. The control method for a robot device according to claim 7, wherein the change is performed.