JP4572080B2 - Robot or behavior update method in robot, and program for controlling behavior update in robot - Google Patents

Description

  The present invention relates to a robot that updates behavior in a robot that communicates with humans, or a device having an interface similar thereto, a behavior update method in a robot, and a program that controls behavior updates in a robot.

  In recent years, robots that communicate with users such as pet robots have been commercialized, and interactive robots have been actively developed.

In many of these interactive robots, data defining the behavior of the robot (hereinafter referred to as behavior data) is stored in advance in an internal storage area provided in the robot, and operates according to the behavior data. However, the capacity of the internal storage area inside the robot is limited, and the behavior pattern of the robot is naturally limited. Therefore, most of the interactive robots repeatedly perform the same action when used for a long period of time, resulting in a problem that the user gets bored.
In addition, since each user has a different preference, there arises a problem that only the action data stored in advance in the storage area cannot sufficiently answer the requests of all users.

  Therefore, in order to solve these problems, a method for dynamically updating behavior data existing in the internal storage area provided in the robot is necessary from an external storage device having a large capacity storage area on the network. In response, the user acquires behavior data, and the user updates the behavior data in the internal storage area of the robot. One example thereof is described in Patent Document 1. By using this method, since the expression of the robot's already appearing action is reduced, it becomes possible to alleviate the user's tiredness.

JP2003-255991

In order to be used by a user for a long time in an interactive robot that mainly communicates with humans, it is important that the user is not bored or uncomfortable. For example, frequent behavior data causes users to get bored. Moreover, the action data which repeats failure causes discomfort to the user. In addition, behavior data that is hardly expressed indicates that the internal storage area is not effectively used, and other behavior data frequently appears, leading to boredom. Therefore, in order to eliminate the user's tiredness and discomfort, unfavorable expression patterns that do not give a very good impression to the user, such as frequent and repeated failures, almost no expression It is necessary to reduce the behavior data characteristic of (hereinafter referred to as undesirable behavior patterns).
However, the method for updating the behavior data of the robot via the network described in Patent Document 1 updates the behavior data so as not to repeat the same behavior, but specifies the factors that cause an undesirable behavior pattern. Therefore, there is a problem that there is a high possibility that the replaced data will continue to appear frequently even if the frequently appearing behavior data is replaced.
Many interactive robots change behavior by selecting behavior data according to external stimuli such as voice recognition and image recognition results and sensor input. Depending on the environment in which the robot is used, even if it is the same robot, there will be a difference in the behavior that appears. For this reason, the behavior data that causes the user to get bored or uncomfortable is also different for each robot, and the countermeasure for reducing the behavior data that causes the tired or unpleasant is also different for each robot.
Therefore, in order to reduce undesirable behavior patterns, there is a need for a method of detecting behavior that causes different undesirable behavior patterns for each robot and updating behavior data.

  In addition, there are types of behavior data such as robot functions, gestures, utterances, and operation parameters. However, if the behavior data is limited to specific types of behavior data, effective behavior data cannot be updated. For example, as a coping method when a frequent utterance line is detected, a method of replacing a function that calls the utterance line as well as replacing it with a similar utterance line can be considered. Moreover, as a countermeasure when a function with many failures is detected, it can be considered to replace it with a higher-performance similar function or to replace an operation parameter used by a function with many failures. In addition, when a gesture that is hardly expressed is detected, simply replacing it with similar data does not provide a solution because the similar data still does not appear. In other words, the behavior data having a characteristic expression pattern and the behavior data to be updated for improvement are not always the same type. Therefore, a flexible update method that does not limit the types of behavior data is required.

  When the user is provided with a function for updating such robot behavior data, the user needs to determine when to update which behavior data. However, in many cases, the user does not have the expertise of the robot, so it is difficult to know the internal state of the robot, it is necessary to update, and if it is necessary to update which behavior data I can't judge what to do. In addition, since it is difficult to determine the similarity of behavior data, when replacing behavior data, it is impossible to determine which behavior data is used to maintain the consistency of behavior before and after the update. Therefore, there is a possibility that the behavior data is updated but is not working effectively, or the consistency of the behavior of the robot is lost and deteriorated. Therefore, the behavior data that is desired to be updated is detected and the behavior data is presented to the user to assist the user in updating the behavior data, or everything from the detection of the behavior data that is desirably updated to the update of the behavior data. A function to perform automatically is required. There is also a need for an update method that does not lose the consistency of the robot behavior before and after the update.

  Therefore, the present invention improves them from undesirable behavioral patterns such as behavioral data that gives users fatigue, such as frequent appearances and rare expression, and behavioral data that makes users uncomfortable, such as repeated failures. To detect behavior data that needs to be updated and maintain the consistency of the behavior before and after the update, update the behavior data to make the robot do not get bored or uncomfortable for the user An object of the present invention is to provide a robot, a behavior updating method for a robot, and a program for controlling behavior updating in the robot.

  The robot according to the present invention uses the behavior data and behavior history stored in the storage device to classify the behavior data into behavior data having the same effect and the same meaning (hereinafter, similar data) and classify the behavior data. A feature behavior detection means that handles behavior data for each unit (hereinafter referred to as behavior type), a search using the behavior type as a key, a semiconductor memory connected to the robot, a magnetic disk, an optical disk, a server connected via a network, etc. Characteristic behavior search means for acquiring a data group belonging to the same behavior type from an external storage device.

  In addition, the behavior data obtained through the behavior search means and the behavior history stored in the storage device are analyzed, and the behavior type that is desirably updated to improve them is detected from the behavior patterns that are not desirably updated. 3 means.

  According to the present invention, the characteristic behavior detection means can handle the behavior data without limiting the type. In other words, action data is not a single type, but includes various types of data such as functions, gestures, utterances, and behavior parameters, but it is related to these types by having a common value called an action type. All can be handled in the same way.

  According to the present invention, when the robot behavior data is updated by the behavior search means, the behavior type is searched as a key, and a data group belonging to the same behavior is acquired from the external storage device and replaced or added. Thus, it is possible to reliably handle similar behavior data, so that the behavior data can be updated while maintaining the consistency of the behavior of the robot before and after the update.

  According to the present invention, the behavior update means analyzes the behavior history of the robot and detects the behavior type that is desired to be updated in order to improve it as an undesired behavior pattern. It becomes possible to cope with. In addition, by detecting an action type that is desired to be updated, the user is required to update the data, and an action type that is desired to be updated is presented, and the user is prompted to update the action data, and the user is burdened. The behavior data belonging to the detected behavior type can be automatically acquired and updated.

  As described above, the robot behavior data update method of the present invention analyzes the behavior history of the robot, and the behavior data that makes the user bored, such as being frequent and rarely expressed, or the user who repeats failure, etc. The behavior data that is unpleasant to the behavior data is detected, and the behavior types that are desired to be updated are detected to improve them, and the data group belonging to the same behavior is acquired from the external storage device, replaced, or added, and the user is bored. Can alleviate and eliminate emotions such as discomfort.

  Next, a first embodiment of the present invention will be described in detail with reference to the drawings.

  Referring to FIG. 1, in the first embodiment of the present invention, a robot 1 that communicates with a person that operates according to behavior data stored in advance in a storage device 11 that is an internal storage area, and that is directly connected to the robot 1. It is configured by an external storage device 2 such as a magnetic disk that can be connected.

The robot 1 includes a storage device 11 that stores information, a behavior executing unit 12, a characteristic behavior detecting unit 13, a behavior detecting unit 14, and a behavior updating unit 15.
The storage device 11 includes an action data storage unit 111 that stores action data of the robot 1 and an action history storage unit 112 that stores an action history of the robot 1.
The external storage device 2 includes a behavior data storage unit 21 that stores robot behavior data.

  In FIG. 1, an arrow is drawn between the robot 1 and the external storage device 2. However, the external storage device may be connected via an existing network such as the Internet. Is a magnetic storage device provided in a personal computer, which downloads behavior data from the Internet or the like via the personal computer, stores the downloaded behavior data in a semiconductor storage device such as a compact flash (R), It is also possible to access the behavior search means provided in the robot 1 via the semiconductor memory device.

Each of these means generally operates as follows.
The behavior executing means 12 reads the behavior data stored in the behavior data storage unit 111 and starts the behavior of the robot 1 according to the read behavior data. In addition, the behavior history storage unit 112 stores which behavior has been executed (which behavior data has been used) and the execution result of the behavior (use data usage information).
The characteristic behavior detection unit 13 analyzes the history of the behavior history storage unit 112 and detects a behavior type that is desired to be updated. The behavior search means 14 receives a behavior type as a search key, and retrieves behavior data belonging to the behavior type received from the behavior data storage unit 21 of the external storage device 2. The behavior update unit 15 receives a behavior data group for updating, and updates the behavior data stored in the behavior data storage unit 111 of the robot 1 using the received data group.
Referring to FIG. 4 and FIG. 6, the behavior data storage unit 111 of the robot 1 and the data of the behavior data storage unit 21 of the external storage device 2 have three values: data ID, behavior data, and behavior type. Is a record, and there are a plurality of records. The data ID value handles a number for uniquely identifying behavior data. The behavior data value handles data defining the robot's behavior such as the robot's function and mechanism, utterance speech, and operation parameters. Since the behavior data value handles data defining the behavior of the robot, the type of data handled varies depending on the configuration of the robot. For example, utterance speech data is unnecessary for a robot that does not utter, such as a dog-type robot, but utterance speech is necessary for a robot that has a speaker and utters. The action type value handles a classification name obtained by finely classifying action data for each similar data, and stores an action type to which the action data belongs. For example, in the case of an utterance line, utterance lines such as “Good morning”, “Good morning” and “Good morning” are classified into action types “greeting_morning_line”. Also, included in the "Hello" in the "greeting _ day _ serif", "Good evening" is "greeting _ night _ dialogue".

  Referring to FIG. 5, the data of the action history unit 112 has a format in which a plurality of records are present with four values of time stamp, action type, data ID, and execution result as one record. The time stamp value stores the time when the action is manifested. The data ID value stores the data ID of the used behavior data, and stores the behavior type to which the behavior data belongs in the behavior type value. The execution result value stores whether the execution of the action data has succeeded or failed. The criteria for success or failure differ for each action data. For example, in the case of a function, if it ends normally, it will be a success, and if it does not end normally, it will be a failure. In the case of an utterance line, a user's reaction after the utterance is acquired using a sensor such as a camera or a microphone. If the user reacts such as being surprised or confused, it is determined to be a failure, and otherwise it is determined to be a success.

  Next, the operation of the action execution unit 12 included in the first embodiment will be described in detail with reference to the flowcharts of FIGS. 1 and 2. This flow starts when the robot 1 is activated. The behavior executing means 12 first reads robot behavior data from the behavior data storage unit 111 (step A11 in FIG. 2). Next, the robot is executed according to the read action data (step A12 in FIG. 2). Next, the data ID of the used behavior data, the behavior type to which the behavior data belongs, the execution start time, and the execution result are stored in the behavior history storage unit 112 as a history (step A13 in FIG. 2). Finally, it is determined whether there is action data to be executed next (step A14 in FIG. 2). If there is action data, the process is repeated from step A1, and if not, the execution of the robot is terminated. Moreover, in the action execution of step A2, when calling another action data among the action data currently being executed, steps A11, A12, and A13 are recursively performed. For example, utterance lines and gestures are mainly called from other action data.

Next, operations of the characteristic behavior detection unit 13, the behavior acquisition unit 14, and the behavior update unit 15 included in the first embodiment will be described with reference to the flowcharts of FIGS. This flow is periodically executed when the robot 1 is activated and the external storage device 2 is connected. It is also executed in response to a user request. First, the characteristic behavior detection means 13 analyzes the information stored in the behavior data storage unit 111 and the behavior history storage unit 112 by paying attention to the behavior data and the behavior type, and whether there is an undesirable behavior pattern. In some cases, an action type that is preferably updated to cope with them is detected (step A21 in FIG. 3). For example, when behavior data that repeatedly fails is detected, the cause of the failure is that there is a problem in the behavior data itself, or that there is a problem in another behavior data used by behavior data such as motion parameters. Etc. are considered. Therefore, as countermeasures for problems, there are methods to replace behavior data that repeatedly fails with other similar data, or to replace other behavior data that is used by behavior data that repeatedly fails. The means 13 uses the action type to which the action data to which the action data that repeatedly fails belongs, or the action type to which the action data that is used by the action data that repeatedly repeats failure belongs. If frequent action data is called from specific action data, add similar data of frequent action data to distribute appearance frequency, or call action data as similar data. Since it is conceivable to suppress the appearance frequency by replacing, the characteristic behavior detection means 13 is the behavior type to which the frequent behavior data belongs or the behavior to which the behavior data calling the frequent behavior data belongs. The type is the detection result.
Next, if an action type that is desired to be updated is not detected, the process ends. If detected, the action type is passed to the action search means 14 (step A22 in FIG. 3). Next, the behavior search means 14 searches the behavior data storage unit 21 of the external storage device 2 for behavior data belonging to the behavior type using the received behavior type as a retrieval key (step A23 in FIG. 3). ). The behavior data group of the search result is passed to the behavior update unit 15. Next, the behavior update unit 15 determines whether there is new data in the received behavior data group (step A24 in FIG. 3). If there is no new behavior data, the process ends. If there is new action data, the data in the action data storage unit 111 of the robot 1 is replaced or added (step A25 in FIG. 3), and the process ends.

  In the first embodiment, the records stored in the behavior data storage unit 111 and the general behavior data storage unit 211 have the behavior data data ID and the behavior type to which the behavior data belongs. It is possible to handle behavior data without distinction of types using types as keys. In addition, since the characteristic behavior detecting means 13 is provided, it is possible to analyze the history and deal with the problems unique to the robot. Further, since the behavior search means 22 is provided, similar data can be reliably searched by searching using the behavior type as a key.

  As an application of the first embodiment, a configuration in which the flow from step A21 to A25 is stored in the behavior data storage unit 111 as one function is also conceivable. At that time, the action data update function is executed in step A12 of FIG. In addition, a configuration in which a plurality of detecting means are stored in the action data storage unit 111 with the characteristic action detecting means 13 as one action type is also conceivable. By adopting such a configuration, the detection means can be made simple by making each detection means simple, such as targeting only frequently spoken speech lines, and only failing functions. Complexity is avoided. Further, when a new type of behavior data is added, it is possible to cope with a change in the behavior data of the robot by adding a detection unit corresponding to the added behavior data.

  As another application, in step A22, the user may be prompted to update the behavior data by indicating that there is an action type that is desired to be updated, or the behavior data update may be confirmed. For example, when an action type that is desired to be updated is detected, the action type that is desired to be updated is presented to the user using a display on a display or an audio output from a speaker, and confirmation is made as to whether or not to perform the update. Upon receiving a response from the user such as input from the keyboard or mouse and voice input from the microphone, if the user's response is affirmative, the process proceeds to step A23, and if not, the process ends. With such a configuration, the user can grasp what kind of behavior data is updated. In addition, it is possible to prevent the behavior data not desired by the user from being updated.

  As another application, when classifying behavior data, a configuration is also considered in which behavior data having the same meaning and the same effect is classified in consideration of the difficulty and complexity of the behavior data. It is done. For example, in an action type of “self-speaking” that talks to the user from the robot, “self-speaking (young)” that is spoken by a child about 4 to 5 years old or “self-speaking” that a child about 6 to 7 years old speaks (Small low) ”,“ Self-sustained utterance (small high) ”, etc., spoken by children about 9-10 years old. With this configuration, for example, each time you confirm that the behavior data belonging to “independent utterance (young)” frequently appears, letting the robot store only “independent utterance (young)” at the beginning. The corresponding data is sequentially replaced with action data belonging to “independent utterances (small and low)”. In addition, every time it is confirmed that the action data belonging to “independent utterance (small and high)” appears frequently, the corresponding data is sequentially replaced with action data belonging to “independent utterance (small and high)”. Can change its behavior as if it grows up. In addition, by replacing frequently used functions from simple functions to complex functions, and replacing rarely used functions from complex functions to simple functions, it is possible to express the skill level of the robot. .

  In describing the best mode for carrying out the present invention, the block diagram shown in FIG. 1 and the flowcharts shown in FIGS. 2 and 3 have been used. However, the present invention can be implemented by a program. Needless to say. In that case, it mainly operates as a program for determining the behavior of the robot. With the configuration of FIG. 1, the configuration of the feature behavior detection means 13, behavior search means 14, behavior update means 15 and the like is generated. I just need it.

  Next, the operation of the present invention will be described using specific examples.

  As shown in FIG. 1, a robot 1 having a recognition device such as a camera, a microphone, an infrared sensor, etc., capable of outputting sound from the microphone or externally outputting to a connected monitor, and communicating with a person. In the external storage device 2 such as a directly connected magnetic disk, the data shown in FIG. 4 is stored in the behavior data storage unit 111 of the robot 1, and the data shown in FIG. 5 is stored in the behavior history storage unit 112. Assume that the behavior data storage unit 21 has the data shown in FIG.

  As a result of analyzing the data in the behavior history storage unit 112 and the behavior data storage unit 111 by the characteristic behavior detection unit 13, when it is found that the speech line “Good morning” appears frequently, to add similar data “Good morning” is the action type to which “good morning” belongs, and the detection result is passed to the action search means 14. The behavior retrieval means 14 retrieves behavior data belonging to “greeting_morning_serif” from the behavior data storage unit 21 of the external storage device 2 using the received “greeting_morning_serif” as a search key, and the retrieval result And “Good morning”, “Good morning” and “Good morning” utterances. The acquired utterance speech is passed to the behavior update means 15. The behavior update unit 15 checks whether there is new data that is not stored in the behavior data storage unit 111 of the robot in the received behavior data group, and utterances of the new data “Good morning” and “Good morning” Is added.

  The present invention can be applied to applications such as a program for updating internal data using data in an external storage device in a robot that communicates with a person and a device having an interface similar thereto. Further, the function of analyzing behavior history and updating problematic behavior data can be applied to applications such as a program for function expansion in a device having communication means with an external storage such as a network home appliance. In addition, the ability to update behavioral data such as utterance lines without breaking the consistency of behavior can be applied to applications such as changing the dialogue and growing thoughts in virtual characters on software agents and games. is there.

It is a block diagram which shows the structure of the 1st Embodiment of this invention. It is a flowchart which shows operation | movement of the action execution means 12 contained in the robot 1 of the 1st Embodiment of this invention. It is a flowchart which shows the operation | movement of the characteristic action detection means 13, the action update means 14, and the server 2 which are contained in the robot 1 of the 1st Embodiment of this invention. It is a figure which shows the data of the action data memory | storage part 111 contained in the robot 1 of the 1st Embodiment of this invention. It is a figure which shows the data of the action history memory | storage part 112 contained in the robot 1 of the 1st Embodiment of this invention. It is a figure which shows the data of the comprehensive action data storage part 211 contained in the server 2 of the 1st Embodiment of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Robot 11 Storage device 111 Behavior data storage device 112 Behavior history storage device 12 Behavior execution means 13 Feature behavior detection means 14 Behavior search means 15 Behavior update means 2 External storage device 21 Behavior data storage section

Claims (9)

A robot that operates according to behavior data that defines behavior stored in a storage area, and the storage area classifies a set of behavior data as behavior types based on meaning and effects, and the behavior data and the data that is the behavior type. And the history of the used behavior data is stored, and the information stored in the history is analyzed by paying attention to the frequency of failure for each behavior data and behavior type. A robot characterized by detecting to which action type frequently- behaved action data belongs, and updating the action data in the storage area according to the detected action type. First behavior data storage means for classifying and storing a set of behavior data as behavior types based on meaning and effect;
Action history storage means for storing a history of used action data;
The behavior data and behavior history stored in the behavior data storage means and the behavior history storage means are analyzed focusing on the failure frequency for each behavior data and behavior type, and the behavior data stored in the behavior data storage means In addition, there is provided characteristic behavior detecting means for detecting a behavior type to which behavior data which is frequently updated and which is behavior data having a high frequency of failure belongs, and the behavior data for which the storage area is desirably updated is changed according to the detected behavior type. A robot characterized by updating to similar behavior data . Separately from the first behavior data storage means, second behavior data storage means for classifying and storing behavior data;
Using the action type of the classified action data as a search key, action search means for searching action data belonging to the action type as a key from the second action data storage means, and action data acquired by the action search means are used. The robot according to claim 2, further comprising behavior update means for updating behavior data stored in the behavior data storage means . First behavior data storage means for classifying and storing behavior data based on meaning and effect;
Second behavior data storage means for classifying and storing behavior data based on meaning and effect;
Action history storage means for storing a history of used action data;
Analyzing the behavior data and behavior history stored in the first behavior data storage means and the behavior history storage means by focusing on the failure frequency for each behavior data and behavior history , Characteristic behavior detection means for detecting which behavior type the behavior data desired to be updated belongs to;
Action search means for searching action data belonging to a key action type from the second action data storage means using the action type of action data detected by the characteristic action detecting means as a search key;
A robot comprising behavior updating means for updating behavior data for which the storage area is desirably updated to another similar behavior data in accordance with the detected behavior type. The robot according to any one of claims 1 to 4, wherein the behavior data includes a plurality of types, and the detection is performed without limiting the types of the behavior data . A method for updating behavior data of a robot that operates according to behavior data that defines the behavior of the robot stored in a storage area provided for the robot,
A set of action data based on meaning and effect is classified and stored as an action type, a history of the used action data is stored, and the history is stored in the action data and the frequency of failure for each action history. , Analyze the behavior type of behavior data with frequent failures , detect behavior types that should be updated ,
Using the detected action type as a search key, search is made for action data stored in a storage area other than the storage area provided in the robot, and the action data for which the storage area is desired to be updated according to the detected action type is changed to another similar action data. how to update behavior data and updates to. The update method according to claim 6, wherein the behavior data includes a plurality of types, and the detection is performed without limiting the types of behavior data . An update program for behavior data of a robot that operates according to behavior data that defines the behavior of the robot stored in a storage area provided for the robot,
Based on the meaning and effect, the set of action data is classified and stored as an action type, and the history of the used action data is stored, and the history is focused on the frequency of failure for each action data and action history. Analyzing and detecting an action type of action data that is frequently updated and is desired action data ,
Searching for action data stored outside the storage area provided in the robot with the detected action type as a search key;
A behavior data update program comprising a step of updating behavior data for which the storage area is desirably updated to another similar behavior data in accordance with the detected behavior type. The update program according to claim 8, wherein the behavior data includes a plurality of types, and the detection is performed without limiting the types of behavior data .

Images