JP2013208681A - Interaction device and interaction control program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To realize purposeful interaction while maintaining a communication atmosphere with a person without using a fixed scenario.SOLUTION: An interaction device includes: an output device outputting a plurality of behaviors; a detection part detecting a state relating to the interaction between a person and the output device; a database registering behavior definition information including a trigger condition showing a condition that outputs the behavior in the output device and category information showing a category of state change of the person that can be generated by the output of the behavior; a retrieval section retrieving, from the database as a behavior candidate, behavior definition information that satisfies the trigger condition by the state detected by the detection part according to an instruction specifying the purpose of approach to the person, the behavior definition information including the category information showing the state change of a predetermined category; a selection part selecting one from the behavior candidates; and a control part executing the behavior corresponding to the selected behavior definition information by the selection part.

Description

  The present disclosure relates to an interaction device and an interaction control program that perform communication in a language and non-language with a person.

  A communication robot, which is one of interaction devices, has a function of performing communication with a person by executing various behaviors including at least one of voice and gestures such as gestures. Techniques have been proposed in which such communication robots are used for watching, for example, elderly people.

  In the process of communication robots interacting with people, a method that combines multiple behaviors according to scenarios that assume various situations, for example, has been proposed as a way to execute behaviors that match the situation that changes from moment to moment. (See Patent Document 1).

Japanese Patent Laid-Open No. 2002-120174

  According to the technique of Patent Document 1 described above, a scenario is created for an interaction service in which a communication robot that imitates the form of a living creature behaves like a living thing, and has a purposeful approach to a person to be watched over. Is possible.

  However, when the communication robot operates based on a scenario created in advance, the behavior executed by the communication robot will be the same if an attempt is made to execute services similar to each other in a situation similar to each other. If the communication robot repeats such uniform behavior, the communication robot itself may be bored with the person being watched over. In that case, there is a possibility that the interaction that is the interaction between the person to be watched over and the communication robot may not be realized.

  In order to distinguish multiple similar situations from each other and to change the behavior performed by the communication robot, if a precise scenario including a huge number of cases is created in advance, the above-mentioned uniformization is avoided. It is possible. However, it is very difficult to create such a detailed scenario.

  An object of the present disclosure is to provide an interaction device and an interaction control program capable of continuously realizing a purposeful action while maintaining an atmosphere of interaction with a person without using a fixed scenario.

  An interaction device according to one aspect detects an output device that outputs a plurality of behaviors including at least one of sound and action, and a situation related to an interaction between a person who is acted on by the output device and the output device. Corresponding to each behavior, a trigger condition that indicates a condition for causing the output device to output the behavior, and type information that indicates a type of state change of the person that can be extracted by the output of the behavior The trigger condition is determined according to a situation detected by the detection unit in response to an instruction that specifies at least one occurrence of the plurality of types of state changes as a purpose of acting on the person. The type of behavior definition information that is satisfied and that indicates the state change of the specified type A search unit that searches the database for behavior definition information including information, a selection unit that selects one of the behavior candidates, and a behavior corresponding to the behavior definition information selected by the selection unit And a control unit that causes the output device to execute.

  An interaction control program according to another aspect detects a situation related to an interaction between an output device and a person subjected to an action performed by outputting a behavior including at least one of voice and action. Then, corresponding to each behavior that can be output by the output device, the trigger condition indicating the condition for causing the output device to output the behavior, and the type of state change of the person that can be extracted by the output of the behavior From the database for registering the behavior definition information including the type information to be indicated, depending on the situation detected by the detection unit in response to an instruction specifying at least one occurrence of the plurality of types of state changes as the purpose of acting on the person Behavior definition information that satisfies the trigger condition, and the specified information Search the behavior definition information including the type information indicating the status change of the type, select one of the behavior definition information obtained as a search result, and send the behavior corresponding to the selected behavior definition information to the output device. Causes the computer to execute the process to be executed.

  According to the interaction device and the interaction control program of the present disclosure, it is possible to continuously achieve a purposeful action while maintaining an atmosphere of exchange with a person without using a fixed scenario.

It is a figure which shows one Embodiment of an interaction apparatus. It is a figure which shows the example of behavior definition information. It is a figure which shows the example of the behavior which contributes to achievement of the objective. It is a figure which shows another embodiment of an interaction apparatus. It is a figure which shows another embodiment of an interaction apparatus. It is a figure which shows the example of the synthetic | combination expectation degree calculated about each behavior candidate. It is a figure which shows another embodiment of an interaction apparatus. It is a figure which shows the example of a situation table. It is a figure which shows another example of behavior definition information. It is a figure which shows the example of the time change of the interest degree UIL. It is a figure which shows another embodiment of an interaction apparatus. It is a figure which shows the hardware structural example of an interaction apparatus. It is a figure which shows the example of the flowchart of the process which detects a condition. It is a figure which shows the example of the flowchart of the process which controls behavior. It is a figure which shows the example of the flowchart of the process which selects behavior. It is a figure which shows the example of the flowchart of the process which updates an expectation degree.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

  FIG. 1 shows an embodiment of an interaction device. The interaction device 10 illustrated in FIG. 1 acts on the user U in response to an instruction Dir received from a host device (not shown) such as a computer device at another location.

  The interaction device 10 includes an output device 11, a detection unit 12, a database 13, a search unit 14, a selection unit 15, and a control unit 16. The search unit 14 searches the database 13 based on the purpose specified by the instruction Dir and the situation detected by the detection unit 12 as described later. The behavior obtained by the retrieval unit 14 by the retrieval is a candidate of behavior to be executed by the interaction device 10 as an action for the user U for the interaction between the interaction device 10 and the user U. The selection unit 15 selects one of the behavior candidates obtained by the search unit 14 and causes the output device 11 to output the selected behavior via the control unit 16.

  The sensor unit SD illustrated in FIG. 1 includes, for example, a temperature sensor, a humidity sensor, and a brightness sensor. The sensor unit SD includes various elements that indicate in which environment the interaction between the output device 11 and the user U is performed, such as the temperature, humidity, and brightness around the output device 11 and the user U. Get the measured value for.

  The output device 11 has a function of outputting a plurality of behaviors including at least one of voice and operation as an action for the user U. The output device 11 is desirably an entity that is recognized as an autonomous individual by the user U, such as a communication robot that imitates the form of a creature such as a bear. The output device 11 may be a display device that displays a CG (Computer Graphics) character or the like. In the following description, the communication robot is simply referred to as a robot. Further, when the output device 11 is a robot, the output device 11 is equipped with a camera, a microphone, a contact sensor, and the like as observation equipment for observing the state of the user U who is a work partner. desirable.

  The detection unit 12 detects a situation related to the interaction between the output device 11 and the user U. The detection unit 12 obtains, as part of the interaction information indicating the situation regarding the interaction, for example, the sensor unit SD described above together with the measurement value regarding the situation of the user U obtained by the observation device mounted on the output device 11. Obtain each measured value. Moreover, the detection part 12 acquires the information which shows the internal state of this output device 11 from the output device 11 as some interaction information. The internal state of the output device 11 is represented by a set of parameters accumulated in the process of the output device 11 performing individual behavior. For example, when the output device 11 is a robot simulating a living creature, the internal state of the output device 11 depends on a parameter group indicating the individuality and autonomy of the robot, such as the emotion of the robot and the degree of familiarity with the user U. Can be represented.

  The database 13 registers behavior definition information that defines each behavior that can be output by the output device 11, for example, corresponding to an identification number assigned to each behavior. Each behavior definition information includes a trigger condition indicating a condition for causing the output device 11 to output the behavior. Further, each behavior definition information includes type information indicating the type of state change of the user U that can be extracted by the output of the behavior, and an expectation degree indicating the high possibility of occurrence of each state change. .

  Here, the state of the user U is, for example, an interest level indicating a degree of interest in another person, an activity level indicating a level of the user U being active, or an emotional expression of the user U. It can be expressed by a plurality of index values including the emotional level indicating the richness of the image. In addition, the degree of interest that the user U shows to others can be estimated based on, for example, how much the user U is interested in the robot functioning as the output device 11. The state change of the user U appears, for example, as a change in at least one of these index values. That is, the change of each index value including the above-described degree of interest, activity, and emotion is an example of a change in the state of the user U.

  In addition, the designer who designs each behavior to be executed by the output device 11 aims to cause a change in a preferable direction for at least one of the above-described index values by executing each behavior. You can also set the behavior. Note that the change in the index value in the preferred direction is, for example, a change in the direction in which the value of the index value is increased in the above-described activity level, interest level, and emotion level. Therefore, when a behavior designer registers each behavior in the database 13, the behavior value corresponding to an index value name indicating the type of the index value that can be expected to change according to the behavior is used as the corresponding behavior type information. It may be set in the definition information.

  The setting of the type information in the database 13 can also be performed based on experimental results as described later. For example, the number of changes that increase the interest level, activity level, and emotional level described above, and the amount of increase in each index value, etc. By conducting an experiment for examining the index value, it is possible to specify with high accuracy an index value that can be expected to change depending on each behavior. Further, based on such experimental results, it is possible to estimate the high possibility that each index value will increase when the user U is acted upon by executing individual behaviors. In each behavior definition information, the value of the expectation set for each type information may be specified in the range of numerical value “0” to numerical value “1” based on the experimental result as described above, The value predicted by the behavior designer based on experience may be set.

  Further, in each behavior definition information, an expectation level indicating a high possibility of changing each index value by executing the behavior may be set corresponding to each index value indicating the state of the user U. In this case, when the degree of expectation is “0”, it indicates that there is no possibility of changing the corresponding index value, and as the degree of expectation approaches the numerical value “1”, the corresponding index value changes with higher accuracy. Indicates to do.

  Further, when the purpose specified by the above-described instruction Dir is a change in the state of the user U indicated by a change in each index value including the degree of interest, the degree of activity, and the degree of emotion described above in a preferable direction, the above-described case is described. The type information indicates the purpose that the individual behavior can achieve. In each behavior definition information, the expectation level corresponding to each index value indicates the degree to which the purpose for each index value is achieved by executing the behavior.

  FIG. 2 shows an example of behavior definition information. The behavior definition information illustrated in FIG. 2 is behavior definition information regarding the behavior specified by the identification number “13” and the behavior name “fresh greeting”.

  This behavior definition information includes a behavior message “Def13” indicating the voice message output by the output device 11 shown in FIG. 1 and the operation as the behavior “refreshing greeting”. The behavior content “Def13” includes, for example, the content of the voice message or the file name of the voice data, and information describing individual operations for realizing gestures.

  Further, the behavior definition information includes a trigger condition “C1 AND C2 AND C3” that specifies a trigger for outputting the behavior “refreshing greeting”. Each of the symbols C1, C2, and C3 shown in FIG. 2 indicates, for example, a range of parameter values that can be detected by the detection unit 12 shown in FIG. 1, and the operator “AND” performs an AND operation. Show. As described above, the trigger condition corresponding to each behavior can be expressed as a logical expression indicating a relationship between one parameter range or a plurality of parameter ranges. According to such a logical expression, for example, “the user U is in a position where the user can act from the output device 11, the discomfort index calculated from the temperature and humidity is equal to or lower than the threshold value, and the ambient brightness is equal to or higher than the threshold value. Complex trigger conditions can be shown as when

  Furthermore, the behavior definition information shown in FIG. 2 includes the types of state changes of the user U that can be extracted by the behavior “refreshing greeting” and the expectation level corresponding to each. In FIG. 2, symbol AIL indicates the degree of expectation for achievement of the state change “increases the activity level of the user U”, and symbol EIL indicates the achievement of the state change “increase the emotion level of the user U”. Indicates the expected degree of. Reference numeral IIL indicates the degree of expectation for achieving the state change “increase the degree of interest of the user U”.

  Each behavior definition information may include an expectation degree regarding other kinds of state changes in addition to the above three kinds of expectation degrees AIL, EIL, and IIL. The value of the expectation degree for each purpose can be set within the range of numerical value “0” to numerical value “1”, for example, when each behavior is registered in the database 13. In the following description, when various expectations including the three types of expectations AIL, EIL, and IIL are not distinguished from each other, they are referred to as expectations XIL.

  The behavior “refreshing greeting” shown in FIG. 2 has the possibility of changing the activity level, emotion level, and interest level of the user U, and therefore the numerical value “0” for each of the expectations AIL, EIL, and IIL. Values other than “0.3”, “0.5”, and “0.3” are set. The database 13 includes a behavior in which a value other than “0” is set as the degree of expectation for a plurality of purposes, as in the example of FIG. 2, while the degree of expectation for all purposes is “0”. May be included. In addition, the database 13 may include behavior definition information in which the degree of expectation is not set for some or all purposes. Note that, in the behavior definition information, when the expectation level is not set for some or all purposes, it can be handled that the expectation level for those purposes is the value “0”.

  Each behavior definition information may include a priority for execution of each behavior as shown in FIG. For example, when each behavior is registered in the database 13, one of a plurality of priorities may be set according to the property of each behavior. For example, for a behavior that is continuously executed, such as a breathing motion, it is desirable to set a low priority in order to preferentially execute another behavior. On the contrary, it is desirable that the behavior for reacting to the action from the user U is executed with priority over the continuous behavior by setting a high priority. In the example of FIG. 2, the medium priority among the three levels of priority “high”, “medium”, and “low” is set for the behavior “fresh greeting”.

  In the interaction device 10 shown in FIG. 1, the search unit 14 searches for behavior candidates from the database 13 in which each behavior as described above is registered as described below.

  For example, the search unit 14 matches the interaction information collected by the detection unit 12 with the trigger condition included in the behavior definition information registered in the database 13 corresponding to each behavior, thereby satisfying the trigger condition. Determine the behavior. Any of the behaviors determined in this way is a behavior that is suitable for the output device 11 to execute in the current situation indicated by the interaction conditions collected by the detection unit 12. Further, the search unit 14 refers to the behavior definition information corresponding to each behavior determined in this manner, and corresponds to the index value name indicating the type of state change designated as the purpose in the instruction Dir, with a numerical value “0”. Detects behaviors with expectation values other than. For example, the search unit 14 corresponds to the degree of interest of the behavior definition information of each behavior determined based on the matching result when the state change “increase the degree of interest” is designated as an action purpose in the instruction Dir. Refer to expectations. Then, the search unit 14 extracts, as one of the behavior candidates, a behavior corresponding to the behavior definition information for which the expectation level having a value other than the numerical value “0” is set. In addition, when extracting the behavior candidate, the search unit 14 sets a predetermined threshold for the expectation corresponding to the index value specified by the instruction Dir, and an extraction condition is that the expectation is equal to or greater than this threshold. Also good. In addition, the search unit 14 first performs extraction based on the expectation corresponding to the purpose specified by the instruction Dir, and the trigger condition set in the behavior definition information corresponding to each behavior obtained by this extraction, Extraction may be performed based on matching with the interaction information.

  The selection unit 15 shown in FIG. 1 selects one of the behavior candidates retrieved from the database 13 in this way. The selection unit 15 may randomly select one behavior candidate from among a plurality of behavior candidates, or may determine a behavior candidate to be selected using a technique such as round robin. Note that a more preferable example will be described later as a method of selecting one from a plurality of behavior candidates in the selection unit 15. The selection unit 15 notifies the selection result by passing the identification number indicating the behavior selected in this way to the control unit 16.

  The control unit 16 shown in FIG. 1 refers to the behavior definition information registered corresponding to the identification number in the database 13 based on the identification number notified from the selection unit 15, and is included in the behavior definition information. Get behavior content. The control unit 16 passes the behavior content to the output device 11 and instructs the output device 11 to execute the behavior content, thereby causing the output device 11 to execute the behavior selected by the selection unit 15. Note that the control unit 16 may perform control for the output device 11 to continuously output, for example, a breathing action for producing a creature-likeness, in addition to the behavior for acting on the user U. Good.

  Here, the behavior candidate obtained by the search unit 14 described above is the purpose specified by the instruction Dir when the trigger condition included in the behavior definition information matches the current situation acquired by the detection unit 12. This is a useful behavior for achieving this. In other words, regardless of which of these behavior candidates is selected, the selected behavior is suitable for the achievement of the purpose specified by the instruction Dir because it matches the current situation. Therefore, the behavior selected by the selection unit 15 is a behavior that can achieve the specified purpose while inheriting the atmosphere of alternating current established by the user U that has been executed by the output device 11 so far.

  Although the number of behavior candidates passed to the selection unit 15 may increase or decrease depending on the situation detected by the detection unit 12, there are a plurality of behavior candidates in many cases. For example, when a robot is used as the output device 11, it is set as an opportunity condition that the line of sight of the robot and the user U matches the behavior definition information corresponding to a large number of behaviors. Then, the search unit 14 determines that the behavior that contributes to the achievement of the purpose of changing each index value from among the behaviors that are set in the behavior definition information in which the opportunity condition “the robot and the user U are in line of sight”. As shown in FIG. 3, a plurality of behavior candidates can be extracted.

  FIG. 3 shows an example of behavior that contributes to the achievement of the objective. Note that among the elements shown in FIG. 3, the same elements as those shown in FIG. 2 are denoted by the same reference numerals, and description thereof is omitted.

  The set of behaviors that contribute to the achievement of the purpose “increase the interest level of the user U” illustrated in FIG. 3 includes the behavior “refreshing greeting” with the identification number 13 and the behavior “refreshing greeting” with the identification number 15. Is included. Furthermore, the set of behaviors that contribute to the achievement of the purpose “increase the degree of interest of the user U” is the behavior of the identification number 22 “I want you to be flashy” and the behavior of the identification number 11 “I am happy with chatter” Including. The example of FIG. 3 indicates that the expectation degree IIL = 0.3 is set in the behavior definition information of the behavior “Refreshing Greeting” in correspondence with the above-described purpose. Similarly, this example shows that the expectation degree IIL = 0.2 is set in the behavior definition information of the behavior “I want to be quiet” corresponding to the above-mentioned purpose. Also, this example shows that the expectation degree IIL = 0.6 is set in the behavior definition information of the behavior “I want to be flashy” corresponding to the above-mentioned purpose. Similarly, this example shows that the expectation degree IIL = 0.7 is set in the behavior definition information of the behavior “Happy and happy” corresponding to the above-described purpose.

  Also, the set of behaviors that contributes to the achievement of the purpose “increasing the activity level of the user U” illustrated in FIG. 3 includes the behavior “Refreshing Greeting” with the identification number 13 and the behavior “Wake a hand” with the identification number 8. Is included. Further, the set of behaviors that contribute to the achievement of the purpose “increase the activity level of the user U” includes the behavior of the identification number 11 “Rejoices” and the behavior of the identification number 34 “Sneezes”. Yes. The example of FIG. 3 indicates that the expectation level AIL = 0.3 is set in the behavior definition information of the behavior “Refreshing Greeting” corresponding to the above-described purpose. Similarly, this example shows that the expectation level AIL = 0.5 is set in the behavior definition information of the behavior “waving hand” corresponding to the above-described purpose. Further, this example shows that the expectation level AIL = 0.8 is set in the behavior definition information of the behavior “Happy and happy” corresponding to the above-described purpose. Similarly, this example shows that the expectation level AIL = 0.8 is set in the behavior definition information of the behavior “sneeze” corresponding to the above-described purpose.

  In addition, the set of behaviors that contribute to the achievement of the purpose “increase the emotion of the user U” illustrated in FIG. 3 includes the behavior “refreshing greeting” of the identification number 13 and the behavior “getting the behavior of the identification number 35”. Is included. Furthermore, the set of behaviors that contribute to the achievement of the purpose “increase the emotional level of the user U” includes the behavior of the identification number 11 “Rejoices”, the behavior of the identification number 121 “says a gag”, and the identification number 8 behaviors “waving hands” are included. Note that the example in FIG. 3 indicates that the degree of expectation EIL = 0.5 is set in the behavior definition information of the behavior “Refreshing Greeting” corresponding to the above-described purpose. Similarly, this example shows that the expectation level EIL = 0.8 is set in the behavior definition information of the behavior “getting” corresponding to the above-described purpose. In addition, this example shows that the expectation level EIL = 0.9 is set in the behavior definition information of the behavior “Happy and happy” corresponding to the above-described purpose. Similarly, this example shows that the degree of expectation EIL = 0.7 is set in the behavior definition information of the behavior “say gag” corresponding to the above-described purpose. Furthermore, this example shows that the expectation degree EIL = 0.6 is set in the behavior definition information of the behavior “waving hand” corresponding to the above-described purpose.

  Note that the symbol UIL shown in FIG. 3 indicates the degree of interest with respect to the robot, which is one of the index values representing the state of the user U. Moreover, the code | symbol UAL shows the activity level which is one of the index values showing the state of the user U, and the code | symbol UEL shows the emotion level similarly. Each of these index values will be described later.

  As can be seen from the example shown in FIG. 3, the set of behavior candidates extracted under the condition that the trigger condition is satisfied by the current situation and contributes to the achievement of the specified purpose is the user U It includes multiple behaviors that perform various actions against the. Therefore, the behavior selected by the selection unit 15 is not fixed to one behavior even if the situation detected by the detection unit 12 and the purpose specified by the instruction Dir are the same, and a set of behavior candidates There is variation in the inside.

  That is, by causing the output device 11 to execute the behavior selected by the selection unit 15 as described above, the behavior executed by the output device 11 can be changed even in the same situation. . The selection unit 15 simply needs to select one from a plurality of behavior candidates, and does not require a complicated branch scenario. On the other hand, the behavior selected by the selection unit 15 is a behavior that can achieve the specified purpose while inheriting the atmosphere of exchange established by the user device U executed by the output device 11 so far. It is as described above. Therefore, the interaction device 10 having the search unit 14 and the selection unit 15 outputs various behaviors that can achieve the specified purpose without impairing the atmosphere of interaction between the user U and the output device 11. It can be output by the device 11.

  As described above, according to the interaction device 10 of the present disclosure, it is possible to continuously realize a purposeful approach while inheriting the atmosphere of exchange established by the approach to the user U without using a fixed scenario. Can do. This is because such an interaction device 10 makes the user U bored by presenting various actions to the user U even when similar objectives are instructed in situations similar to each other. This is because there is little fear. Therefore, according to the interaction device 10 of the present disclosure, it is possible to maintain a natural exchange between the user U and the output device 11.

  The interaction device 10 of the present disclosure is useful for watching over the user U while providing a service that gives the user U a sense of healing using, for example, a robot that imitates the form of a living thing as the output device 11. It is useful in the realization of services that achieve the above.

  For example, for each behavior, by setting an opportunity condition that reflects the ecology of the creature imitated by this robot and an opportunity condition that reflects the characteristics that are unique to each robot, It is possible to work on the user U more naturally.

  FIG. 4 shows another embodiment of the interaction device 10. 4 that are the same as those shown in FIG. 1 are given the same reference numerals, and descriptions thereof are omitted.

  The selection unit 15 illustrated in FIG. 4 includes a calculation unit 151 and a determination unit 152. The calculation unit 151 illustrated in FIG. 4 calculates, for each behavior candidate extracted from the database 13 by the search unit 14, a probability that has a correlation with the high degree of expectation for the type of state change specified by the instruction Dir. . Further, the determination unit 152 determines a behavior to be selected from the behavior candidates according to the probability obtained for each behavior candidate by the calculation unit 151. As described above, the calculation unit 151 and the determination unit 152 may realize weighted random selection in which one of a plurality of behavior candidates is selected according to the probability weighted with the degree of expectation.

  For example, as the probability Pj that the j-th behavior candidate is selected from the N behavior candidates, the calculation unit 151 sets the expectation degree XIL (j) corresponding to the purpose set for each behavior candidate to these expectations. It may be calculated as a value divided by the sum S of degrees. In this case, the probability Pj can be expressed as in Expression (1) using the expectation degree XIL (j) for the designated purpose of each behavior candidate.

指示Ｄｉｒによって目的「利用者Ｕの興味度を上昇させる」が与えられた場合について、図３に示した例に基づいて説明すれば、各ふるまいが選択される確率は、次のように計算することができる。ふるまい「さわやかな挨拶」が選択される確率は、ふるまい「さわやかな挨拶」の期待度０．３を、４つのふるまいに対応する期待度の総和１．８で除算した値０．１８となる。同様にして、ふるまい「地味にかまって欲しがる」が選択される確率は、ふるまい「地味にかまって欲しがる」の期待度０．２を上述した期待度の総和１．８で除算した値０．１１となる。また、ふるまい「派手にかまって欲しがる」が選択される確率は、ふるまい「派手にかまって欲しがる」の期待度０．６を上述した期待度の総和１．８で除算した値０．３３となる。そして、ふるまい「はしゃいで喜ぶ」が選択される確率は、ふるまい「はしゃいで喜ぶ」の期待度０．７を上述した期待度の総和１．８で除算した値０．３９となる。

If the purpose “increase the interest level of the user U” is given by the instruction Dir, based on the example shown in FIG. 3, the probability that each behavior is selected is calculated as follows. be able to. The probability that the behavior “Refreshing Greeting” is selected is 0.18, which is obtained by dividing the expectation level 0.3 of the behavior “Refreshing Greeting” by the sum of 1.8 of the expectation levels corresponding to the four behaviors. Similarly, the probability that the behavior “I want to be quiet” is selected by dividing the expectation level 0.2 of the behavior “I want to be quiet” by the sum of the above expectations of 1.8. The value is 0.11. In addition, the probability that the behavior “I want to be flashy” is selected is a value obtained by dividing the expectation level 0.6 of the behavior “I want to flashy” by the sum of the above-mentioned expectations of 1.8. .33. Then, the probability that the behavior “happy with pleasure” is selected is 0.39, which is obtained by dividing the expectation degree 0.7 of the behavior “happy with pleasure” by the total of the above-mentioned expectation degrees 1.8.

  For example, the determination unit 152 illustrated in FIG. 4 performs a probabilistic trial to generate a random number in a predetermined range, and based on the obtained random number value and the probability determined as described above, The behavior to be selected from the behavior candidates is determined. The determining unit 152 divides the range of the random numbers described above into N sections having a width corresponding to the probability size obtained for each behavior candidate, and which section includes the generated random numbers. Depending on, the behavior to be selected may be determined.

  The selection unit 15 including the calculation unit 151 and the determination unit 152 described above calculates a high degree of expectation that is a high possibility that the purpose specified by the instruction Dir is achieved by each behavior from the behavior candidates. The behavior to be executed is selected by the reflected weighted random selection. Therefore, the selection result by the selection unit 15 shown in FIG. 4 indicates that the degree of expectation regarding achievement of the purpose specified by the instruction Dir while maintaining diversity within the range of behavior candidates obtained by the search by the search unit 14. Reflects the height of.

  According to the interaction device 10 having such a selection unit 15, the behavior that is highly likely to achieve the purpose specified by the instruction Dir is preferentially output while maintaining the variety of behaviors output by the output device 11. Can be made. As a result, the output device 11 can execute an action with a high possibility of achieving the object while avoiding the action that the output device 11 performs to the user U from being uniform.

  Note that the interaction device 10 disclosed herein can cope with a case where a plurality of purposes are designated by the instruction Dir by providing the selection unit 15 as shown in FIG.

  FIG. 5 shows another embodiment of the interaction device 10. Note that among the components shown in FIG. 5, components equivalent to those shown in FIG. 4 are denoted by the same reference numerals, and description thereof is omitted.

  The selection unit 15 illustrated in FIG. 5 includes a counting unit 153 in addition to the calculation unit 151 and the determination unit 152 similar to those in FIG.

  The totaling unit 153 shown in FIG. 5 refers to the database 13, and each of the M purposes specified by the instruction Dir in the behavior definition information corresponding to each behavior candidate obtained as a search result by the search unit 14. Get the expectation corresponding to. In addition, the totaling unit 153 calculates a composite expectation degree indicating the degree of contribution of each behavior candidate to the achievement of the M purposes described above by summing the expectation degree corresponding to each purpose acquired for each behavior candidate. To do.

  For example, the totaling unit 153 may calculate the combined expectation level XILs (j) for the jth behavior candidate among the N behavior candidates using Equation (2). In Equation (2), the combined expectation degree XILs (j) is the product of the expectation degree XIL (m, j) corresponding to the mth objective and the weight c (m) designated for the mth objective. Is the sum of

各目的に対応する重みｃ(ｍ)の値は、指示Ｄｉｒによって指定してもよいし、集計部１５３の内部に、目的ごとに予め設定しておいてもよい。また、集計部１５３は、全ての目的について、同一の重みを設定してもよい。いずれの場合にも、各目的に対応する重みｃ(ｍ)は、それぞれ正の値であり、重みｃ(ｍ)の総和は数値「１」となるように設定することが望ましい。

The value of the weight c (m) corresponding to each purpose may be specified by the instruction Dir, or may be set in advance for each purpose in the counting unit 153. The totaling unit 153 may set the same weight for all purposes. In any case, it is desirable that the weights c (m) corresponding to each purpose are positive values, and the sum of the weights c (m) is set to a numerical value “1”.

For example, the behavior “Refreshing Greeting” shown in FIG. 2 is extracted as the jth behavior candidate in accordance with the two purposes “increased interest” and “increased emotion” specified by the instruction Dir. The synthesis expectation degree XILs (j) at this time can be obtained as shown in Expression (3).
XILs (j) = 0.3 × 0.5 + 0.5 × 0.5 = 0.4 (3)
In addition, the example of Formula (3) has shown the case where the same weight 0.5 is set in the totaling unit 153 for the two purposes described above. Further, the expression (3) is obtained by setting the number of objectives M = 2 in the expression (2) as the first target expected degree XIL (1, j) and the second target expected degree XIL (2, j). This is an example in which the expectation degree IIL = 0.3 and the expectation degree EIL = 0.5 of the behavior “refreshing greeting” are set.

  Similarly, in FIG. 3, when each behavior listed as an example of behavior that contributes to the achievement of the above-described two purposes is extracted as a behavior candidate, the combined expectation degree calculated by the aggregation unit 153 for each behavior Is shown in FIG.

  FIG. 6 shows an example of the synthesis expected degree calculated for each behavior candidate. 6 that are the same as those in FIG. 3 are denoted by the same reference numerals and description thereof is omitted.

  In the example of FIG. 6, the composite expectation level corresponding to the behavior “I want you to be quiet” weights the expectation level IIL = 0.2 set corresponding to the interest level in the behavior definition information of this behavior. A value of 0.1 multiplied by 0.5 is obtained. Similarly, the combined expectation degree corresponding to the behavior “I want to be flashy” is a weight 0.5 to the expectation degree IIL = 0.6 set corresponding to the interest degree in the behavior definition information of this behavior. Multiplied by the value 0.3. On the other hand, the synthesis expectation level corresponding to the behavior “Happy and happy” is the expectation levels IIL = 0.7 and EIL = 0.9 respectively set according to the interest level and the emotion level in the behavior definition information of the behavior. The result is 0.8, which is obtained by multiplying each by a weight of 0.5. Further, the synthesis expectation level corresponding to the behavior “Gep is” is a value obtained by multiplying the expectation level EIL = 0.8 set corresponding to the emotion level by the weight 0.5 in the behavior definition information of the behavior. .4. Similarly, the synthesis expected degree corresponding to the behavior “say gag” is a value obtained by multiplying the expectation degree EIL = 0.7 set corresponding to the emotion degree by the weight 0.5 in the behavior definition information of the behavior. 0.35. The synthesis expectation level corresponding to the behavior “waving hand” is obtained by multiplying the expectation level EIL = 0.6 set corresponding to the emotion level by the weight 0.5 in the behavior definition information of the behavior. The value is 0.3.

  Based on the composite expectation degree XILs (j) obtained in this way, the calculation unit 151 converts the expectation degree X (j) included in the above-described expression (1) into the above-described composite expectation degree XILs (j). By using the replaced expression (4), the probability Pj that each behavior candidate is selected is calculated.

このようにして得られた確率Ｐｊは、各ふるまい候補がＭ個の目的の達成に寄与する度合いを示す合成期待度を反映している。つまり、図５に示した算出部１５１で得られた確率に従って、決定部１５２が、Ｎ個のふるまい候補から選択するふるまいを決定することにより、複数の目的を達成する可能性の高さを考慮してふるまいを選択することができる。

The probability Pj obtained in this way reflects a composite expectation that indicates the degree to which each behavior candidate contributes to the achievement of M objectives. That is, considering the probability obtained by the calculation unit 151 illustrated in FIG. 5, the determination unit 152 determines the behavior to be selected from the N behavior candidates, thereby taking into consideration the high possibility of achieving multiple objectives. You can then select the behavior.

  By the way, the behavior illustrated in FIG. 3 includes one in which the effect due to the execution of the behavior changes depending on the state of the user U who is challenged by the interaction device 10. For example, the behavior effect accompanied by the flashy movement of the robot functioning as the output device 11 such as the behavior “Rejoices in a chat” is a great effect when the user U is interested in the interaction with the robot. Demonstrate. For this reason, in the behavior definition information for behavior that causes the robot to execute such a flashy movement, a degree of expectation having a relatively large value is set corresponding to the purpose of increasing each index value related to the state of the user U. Often done. Therefore, when the behavior to be executed by the robot is selected according to the probability reflecting the degree of expectation, the behavior with the flashy movement as described above is selected with a higher probability than other behaviors. However, when the user U is not very interested in the robot, if the behavior with a flashy movement such as the behavior “Happy and happy” is executed, the expected effect cannot be obtained and the opposite effect is obtained. Sometimes it becomes.

  Therefore, it is desirable that the interaction device 10 has a mechanism for controlling the selection of the behavior to be output as an action according to the state of the user U who is the person to whom the action is made.

  FIG. 7 shows another embodiment of the interaction device 10. 7 that are the same as those shown in FIG. 4 are given the same reference numerals, and descriptions thereof are omitted.

  The interaction device 10 illustrated in FIG. 7 is an example of a configuration that realizes the output device 11 by using a part of the function of the robot CR that imitates the shape of a bear.

  The output device 11 illustrated in FIG. 7 includes a motor unit 31, a speaker 32, and a drive processing unit 33 included in the robot CR. The drive processing unit 33 drives the plurality of motors included in the motor unit 31 based on the behavior content received from the control unit 15 and moves the operating parts such as the limbs and the neck of the robot CR. Make a defined action. In addition, the drive processing unit 33 presents the voice message defined by the behavior content to the user U by causing the speaker 32 to output the voice specified by the behavior content received from the control unit 15.

  The robot CR shown in FIG. 7 includes an observation unit 34 and an internal state determination unit 35 in addition to the motor unit 31, the speaker 32, and the drive processing unit 33 described above.

  The observation unit 34 includes, for example, a camera, a microphone, a contact sensor, and the like, and includes information on surrounding images including the user U, voice of the user U, and physical contact from the user U to the robot CR. Are acquired as information representing the reaction from the user U. Moreover, the observation part 34 may acquire the measured value which shows the magnitude | size of a user's U body movement and the change of a facial expression from the image | video obtained with the camera, for example. Similarly, for example, the observation unit 34 may acquire a measurement value indicating the loudness of the user U, the loudness variation of the voice, and the like from the sound obtained by the microphone. In addition, the observation unit 34 may acquire a measurement value indicating the number of times the user U has physically touched the robot CR, the strength, and the like from information obtained by the contact sensor.

  The internal state determination unit 35 calculates each parameter indicating the internal state of the robot CR based on the behavior content received from the control unit 15 by the drive processing unit 33 and the information obtained by the observation unit 34. For example, the internal state determination unit 35 may obtain a parameter indicating the short-term emotion of the robot CR as one of the parameters indicating the internal state of the robot CR based on the content of each behavior received from the control unit 15. Good. The internal state determination unit 35 may adjust the parameter indicating the short-term emotion of the robot CR based on a reaction from the user U obtained by the observation unit 34. For example, when the internal state determination unit 35 detects that the user U has stroked the robot CR by the contact sensor, the internal state determination unit 35 increases the value of a parameter indicating a happy emotion, and a sad emotion when detecting that the user U has been hit. Adjustments such as increasing a parameter indicating the above may be performed. Further, the internal state determination unit 35 indicates, for example, the mood of the robot CR that changes due to the accumulation of emotion as described above and the degree of familiarity with the user U based on a history of a plurality of behavior contents received in time series. A parameter may be obtained. Further, for example, the internal state determination unit 35 determines whether or not the user U and the robot CR are in an eye contact state in which the eyes meet each other based on the video of the user U received from the observation unit 34. You may obtain | require the parameter to show.

  The detection unit 12 illustrated in FIG. 7 includes a collection unit 121, an analysis unit 122, and a situation table 123 in addition to the observation unit 34 and the internal state determination unit 35 described above.

  The collection unit 121 periodically collects measurement values obtained by, for example, a temperature sensor, a humidity sensor, or a brightness sensor included in the sensor unit SD. In addition, the collection unit 121 stores the collected measurement values in the situation table 123 as environment information indicating the situation regarding the environment around the output device 11 and the user U.

  Further, the analysis unit 122 analyzes each measurement value obtained by the observation unit 34, thereby indicating each index indicating the state of the user U, including the activity level UAL, the emotion level UEL, and the interest level UIL illustrated in FIG. Calculate the value. For example, Japanese Patent Application No. 2011-243573 “Non-Language Information Analysis Device and Non-Language Information Analysis Program” previously filed by the applicant of the present application is used for the processing of calculating the activity level UAL, emotion level UEL, and interest level UIL In addition, a technique of “non-linguistic information analysis method” may be used. In this technique, a plurality of measurement values indicating the strength of the non-verbal reaction expressed by the user U are acquired from information obtained from a camera or a contact sensor included in the observation unit 34, and time variations of these measurement values are obtained. To calculate the activity level UAL, the interest level UIL, and the emotion level UEL.

  The analysis unit 122 receives measurement values indicating the magnitude and frequency of movement of the user U from the observation unit 34 and changes in the facial expression of the user U, and performs analysis processing based on the above-described technique for these measurement values. Do. The analysis unit 122 performs, for example, the above-described analysis processing at predetermined time intervals, so that the index values including the activity level UAL, the emotion level UEL, and the interest level UIL are measured values used for calculating these index values. May be calculated for each sampling time. Further, the analysis unit 122 may cause the status table 123 to hold each index value calculated from the measurement value obtained at each sampling time as user information indicating the state of the user U corresponding to the sampling time. Good.

  For example, as shown in FIG. 8, the situation table 123 includes environment information received from the collection unit 121 and robot information indicating the internal state of the robot CR received from the internal state determination unit 35, together with the above-described user information. Is stored in correspondence with the time of acquisition.

  FIG. 8 shows an example of the situation table 123. Of the elements shown in FIG. 8, the same elements as those shown in FIG. 3 are designated by the same reference numerals, and the description thereof is omitted.

  The situation table 123 shown in FIG. 8 stores environment information including the measurement value Li measured by the brightness sensor corresponding to the time Ti, and includes the measurement value Lj measured by the brightness sensor corresponding to the time Tj. Stores environmental information. In addition, this situation table 123 similarly stores environment information including the measurement value Lp by the brightness sensor corresponding to the time Tp.

  In addition, the situation table 123 stores robot information including eye contact “Yes” and mood “Good mood” corresponding to time Ti, and eye contact “Yes” and “Yes” corresponding to time Tj. Stores robot information including the mood “Gokugen”. On the other hand, this situation table 123 stores robot information including eye contact “none” and mood “blaze” corresponding to the time Tp.

  Further, the situation table 123 shown in FIG. 8 stores user information including an interest level UIL (i), an activity level UAL (i), and an emotion level UEL (i) corresponding to the time Ti. The situation table 123 stores user information including an interest level UIL (j), an activity level UAL (j), and an emotion level UEL (j) corresponding to the time Tj. In the situation table 123 shown in FIG. 5, the sign “−” shown in the user information column corresponding to the time Tp indicates that the analysis unit 122 calculates the interest level, the activity level, and the emotion level at the time Tp. Indicates that it was not done.

  Here, in the technique of Japanese Patent Application No. 2011-243573, the robot CR establishes eye contact with the user U in order to calculate the degree of interest, the degree of activity, and the degree of emotion that faithfully reflects the state of the user U. It is assumed that high-accuracy measurement values obtained in the state of use are used. Therefore, as shown by the robot information corresponding to the time Tp in FIG. 8, when the eye contact is not established, the interest level, the activity level, and the emotion level calculated by the technique of Japanese Patent Application No. 2011-243573 are uncertain. May be a negative value. Instead of storing such an uncertain value, the analysis unit 122 stores a sign “−” indicating that the interest level, the activity level, and the emotion level are not calculated, as in the example illustrated in FIG. 8. May be.

  The status table 123 is not limited to the example illustrated in FIG. 8, and the latest environment information, the latest robot information, and the latest user information are displayed on the status regarding the interaction between the output device 11 and the user U. The latest interaction information may be held.

  As described above, according to the detection unit 12 including the analysis unit 122 illustrated in FIG. 7, each index value indicating the state of the user U is obtained from the situation regarding the interaction between the output device 11 and the user U. It can be passed to the search unit 14 as part of the interaction information to be represented.

  Further, in the interaction device 10 having the detection unit 12 including the analysis unit 122 described above, even when an opportunity condition including a condition indicating a range of each index value indicating the state of the user U is set in each behavior definition information. Good.

  FIG. 9 shows another example of behavior definition information. Note that, among the elements shown in FIG. 9, elements equivalent to those shown in FIG. 2 are denoted by the same reference numerals, and description thereof is omitted.

  The behavior definition information illustrated in FIG. 9 is behavior definition information regarding the behavior specified by the identification number “11” and the behavior name “Happy and happy”.

  This behavior definition information includes a behavior message “Def11” indicating a voice message and an operation output by the output device 11 shown in FIG. This behavior content “Def11” describes, for example, that the limbs of the robot CR shown in FIG. 7 are moved large and violently, and that a voice message with a soft tone is output at a high volume.

  If the user CR is eager to interact with the robot CR and the robot CR executes such a flashy behavior, the user U's degree of interest, activity and emotion can be estimated with high probability. As described above, it can be raised. For this reason, the behavior definition information shown in FIG. 9 has values “0.8” and “0” as the expected degrees AIL, EIL, and IIL of changing the activity level, emotion level, and interest level of the user U, respectively. .9 ”,“ 0.7 ”, etc., which are close to the maximum value“ 1 ”.

  On the other hand, when the user U is not very interested in the interaction with the robot CR, when the robot CR executes the action by the flashy behavior as described above to the user U, the interaction is continued. As mentioned above, there is a risk of making it difficult.

  The behavior definition information shown in FIG. 9 is one of indices indicating the state of the user U as a part of the trigger condition in order to suppress the output of the behavior in a case where such a risk is assumed. A condition indicating a range for the interest level UIL is included.

  In the trigger condition “C1 AND C4” set in the behavior definition information shown in FIG. 9, the condition C4 indicates a condition indicating the range of the interest degree UIL of the user U obtained by the analysis unit 122 shown in FIG. ing. In the example of FIG. 9, the condition C4 is that the degree of interest UIL of the user U with respect to the robot CR exceeds the numerical value “0.8”.

  Based on such a trigger condition, the search unit 14 shown in FIG. 7 searches for a behavior candidate, so that the behavior is limited only when the user U shows a great interest in the interaction with the robot CR. “Happy and happy” can be extracted as one of the behavior candidates.

  As described above, the search unit 14 illustrated in FIG. 7 first collates the latest interaction information included in the situation table 123 with the trigger condition included in each behavior definition information in the database 13. Extract the behavior that satisfies the condition. Next, the search unit 14 achieves the purpose depending on whether or not a value other than “0” is set as the degree of expectation corresponding to the purpose specified by the instruction Dir for each behavior extracted in this way. Valid behaviors are identified as behavior candidates.

  Therefore, the behavior extracted as a behavior candidate by the search unit 14 is limited to a behavior in which the trigger condition is satisfied by the latest interaction information including each index value indicating the state of the user U. In other words, the behavior “Happy and happy” with the trigger condition shown in FIG. 9 shows that the interest level of the user U exceeded the numerical value “0.8” due to the latest interaction information held in the situation table 123. Unless it is shown, it is not extracted as a behavior candidate.

  Here, each index value including the degree of interest UIL of the user U with respect to the robot CR varies with time in accordance with, for example, accumulation of actions from the robot CR as shown in FIG.

  FIG. 10 shows an example of the temporal change in the degree of interest UIL. Note that among the elements shown in FIG. 10, elements equivalent to those shown in FIG. 8 are denoted by the same reference numerals, and description thereof is omitted.

  In the example illustrated in FIG. 10, the degree of interest UIL of the user U with respect to the robot CR generally increases with time. In this interest degree UIL graph, the values of the interest degree UIL at the times Ti and Tj are both less than the numerical value “0.8”. At the subsequent time Tk, the value of the interest degree UIL is the numerical value “0.8”. Has been reached.

  In accordance with such a change in the index value, whether or not the behavior candidate searched by the search unit 14 from the database 13 includes a behavior having a risk such as “being happy” is changed. For example, when the trigger conditions of each behavior shown in FIG. 3 as examples of behavior that contributes to the purpose “increase activity” are all satisfied except for the conditions related to the interest UIL of the user U, the behavior A candidate changes as follows according to the change of the degree of interest UIL.

  When the degree of interest UIL of the user U with respect to the robot CR varies as shown in FIG. 10, the trigger condition for the behavior “pleasantly pleased” described above is not satisfied at time Ti and time Tj before time Tk.

  Therefore, for example, even at the time Ti, even if the purpose specified by the instruction Dir is the purpose contributed by the behavior “Happy and happy”, the search unit 14 shown in FIG. 7 selects the behavior “Happy and happy” as the behavior candidate. Not to extract as. This is because, in the process of collating the interaction information with the trigger condition in each behavior definition information, the search unit 14 determines that the trigger condition is not satisfied and excludes it from the candidate behavior. . In this case, the behavior candidates extracted by the search unit 14 in response to the instruction Dir designating the purpose “increase activity” are the behavior “refreshing greeting”, the behavior “waving hands”, and the behavior “sneezing”. It is three.

  On the other hand, if an instruction Dir that designates one of the purposes to which the behavior “joyful” is input after time Tk shown in FIG. 10, the search unit 14 shown in FIG. "Rejoice" is extracted as one of the behavior candidates. Therefore, after the time Tk, the behavior candidate extracted by the search unit 14 in response to the instruction Dir designating the purpose “increase activity” is the above-mentioned three behaviors plus 4 behaviors “pleasantly rejoice” 4 One.

  As described above, according to the interaction device 10 shown in FIG. 7, the interaction apparatus 10 responds to changes in the interaction state including the state of the user U that is the target of the interaction, without depending on the densely constructed fixed scenario. In this way, it is possible to execute various and varied actions. This reduces the risk that the robot CR will be bothersome for the user U due to the behavior of “being happy”, and the interaction between the robot CR and the user U can be continued stably. Can do.

  Whether or not each behavior registered in the database 13 has a risk as described above is determined by accumulating responses obtained when each behavior is presented to various persons in various situations. This can be determined in advance by experiments. Further, the factors that influence the effect of the behavior can be determined in advance based on the same experimental results.

  And the behavior having the risk determined as described above, as well as the behavior described above, the trigger condition including the condition for the index value indicating the state of the user U that determines the effect of achieving the objective. Can be set. As a result, since these behaviors can be included in the behavior candidates only in situations where these behaviors work effectively for the user U, the risk of outputting the behaviors as described above by the robot CR is reduced. Can be avoided.

  As described above, according to the interaction device 10 shown in FIG. 7, the interaction between the robot CR and the user U is enhanced by setting the trigger condition including the condition related to the state of the user U for a part of the behavior. While continuing with accuracy, it is possible to achieve an action according to the purpose.

  By the way, all the state changes specified as the purpose of acting on the user U by the instruction Dir appear as changes in the respective index values including the interest degree UIL, the activity degree UAL, and the emotion degree UEL. Therefore, the index value obtained as an analysis result of the analysis unit 122 after executing a certain behavior of the robot CR in order to extract a change in the index value specified by the instruction Dir indicates the target achievement level as it is. Therefore, based on the change of each index value obtained by the analysis unit 122, the effect of the behavior presented by the output device 11 to the user U is evaluated according to the designation of the purpose related to the index value, and the evaluation result is obtained. It is possible to optimize the expectation based on this.

  FIG. 11 shows another embodiment of the interaction device 10. 11 that are the same as those shown in FIG. 7 are denoted by the same reference numerals and description thereof is omitted.

  The interaction device 10 shown in FIG. 11 includes an update unit 17 in addition to the output device 11, the detection unit 12, the database 13, the search unit 14, the selection unit 15, and the control unit 16 shown in FIG.

  The update unit 17 determines, based on the change in the index value obtained by the analysis unit 122 before and after the output of the behavior, the type of behavior definition information corresponding to the behavior output by the robot CR, which is designated as the purpose of the action in the instruction Dir. Update the expectation level for the state change.

  For example, the update unit 17 first refers to the situation table 123 at a timing before and after the output device 11 outputs the behavior selected according to the instruction Dir, thereby indicating an achievement level corresponding to the designated purpose. Get the values before and after the behavior output. Next, the update unit 17 subtracts the achievement obtained correspondingly before outputting the behavior from the achievement obtained correspondingly after outputting the behavior described above, thereby changing the achievement due to the output of this behavior. Find the minute δ. The update unit 17 receives, for example, a notification indicating the timing before and after the output device 11 outputs the behavior selected according to the instruction Dir from the control unit 15, and refers to the status table 123 according to this notification. May be.

  For example, when the degree of interest UIL changes as shown in FIG. 10, for the behavior presented during the period from time Ti to time Tj, the update unit 17 changes the achievement degree of the purpose “increase the degree of interest”. As a minute, a change δ of the degree of interest UIL in the above-described period is obtained.

Further, the update unit 17 shown in FIG. 11 performs a new process related to the purpose specified by the instruction Dir in the behavior definition information corresponding to the behavior described above, based on the change δ in the achievement level obtained as described above. Calculate a high degree of expectation. Based on the above-described change δ of the achievement level, the predetermined coefficient k, and the expected level XILb before the execution of the behavior, the update unit 17 performs, for example, a new expectation expressed by the following equation (5) The degree XILn may be calculated.
XILn = XILb + k × δ (5)
The update unit 17 appropriately sets the value of the coefficient k shown in Equation (5), thereby reflecting the learning result based on the reaction of the user U when presenting each behavior in the corresponding expectation. The degree can be adjusted. Note that the update unit 17 sets a relatively small value as the value of the coefficient k described above, and the new expectation is changed to the previous expectation by the change δ of the achievement obtained when each behavior is executed. It is desirable to make adjustments so that it does not change excessively and gradually changes according to daily contact. Further, it is desirable that the updating unit 17 restricts the value of the new expected degree XILn calculated as described above to a range from a numerical value “0” to a numerical value “1” that is a range of values as the expected degree. .

  In addition, the updating unit 17 uses the new expectation degree XILn calculated as described above, for the behavior specified for the behavior definition information registered in the database 13 for the behavior output by the output device 11. Performs processing to rewrite the corresponding expectation.

  In this way, the update unit 17 calculates a new expectation level that reflects the change δ of the index value obtained by the analysis unit 122 before and after the behavior output, and uses this new expectation level to calculate the database 13. By updating the information, the degree of expectation can be optimized.

  In this way, in the process of accumulating the interaction with the user U, the search unit 14 and the selection unit 15 change the expectation that each behavior leads to the change of each index value of the user U. The tendency of the behavior to select according to the designated purpose also changes. This change is a change in a direction in which the search unit 14 and the selection unit 15 select a behavior that works more effectively by the user U in achieving the specified purpose. That is, as described above, the search unit 14 and the selection unit 15 work more effectively on the user U by updating the expectation level corresponding to the update unit 17 according to the change in the situation detected by the detection unit 12. Learning for selecting the behavior can be realized.

  That is, according to the interaction device 10 having the updating unit 17 illustrated in FIG. 11, customization for selecting a behavior that suits the individuality of the user U is realized while continuing to interact with the user U. be able to.

  The customization that can be realized by the interaction device 10 having the updating unit 17 shown in FIG. 11 can cope with the case where the user U gets bored with individual behavior.

  Because, when the user U gets bored with a certain behavior A, the change δ of the index value obtained by the analysis unit 122 based on the reaction that the user U shows before and after the presentation of the behavior A is This is because the possibility of having a negative value is high. As is apparent from the equation (5), when the change δ is a negative value, the value of the new expectation degree XILn is smaller than the expectation degree XILb before the behavior A is executed. Thereby, compared with the probability that the behavior A is selected again, the probability that the other behavior is selected becomes relatively large, and thus it is possible to give a fresh stimulus to the user U.

  The method by which the update unit 17 calculates the new expectation level XILn is not limited to the equation (5), and any other method may be used as long as the index value change δ is reflected in the new expectation level. . For example, for each of the plurality of sections provided for the change amount δ described above, an increment of the expectation degree XIL to be applied when the expectation degree XIL is updated is set in advance, and the change corresponding to the section to which the index value change amount δ belongs A new expectation degree XILn may be calculated by applying an increment of the expectation degree XIL.

  As described above, the interaction device 10 disclosed herein can easily realize automatic customization for each user U to be interacted with. In addition, the interaction device 10 disclosed herein adds a new behavior definition information in which an appropriate trigger condition and an expectation level corresponding to each purpose are set to the database 13, and can be presented to the user U. Can be easily expanded. As a result, it is possible to prevent the occurrence of a situation where the user U gets bored with higher accuracy with respect to the interaction via the robot CR, so that the interaction can be stably maintained over a long period of time. Is possible.

  The interaction device 10 disclosed herein can be realized by using a computer device such as a personal computer, for example.

  FIG. 12 shows an example of the hardware configuration of the image processing apparatus 10. Note that among the components shown in FIG. 12, components equivalent to those shown in FIG. 7 are denoted by the same reference numerals and description thereof is omitted.

  The computer device 20 includes a processor 21, a memory 22, a hard disk device (HDD) 23, a display device 24, an input device 25, an optical drive device 26, and a network interface 28. Further, the computer device 20 includes a peripheral device interface 29 and a robot interface 30. In FIG. 12, the interaction device 10 includes a processor 21, a memory 22, a hard disk device 23, a peripheral device interface 29, a robot interface 30, and a robot CR.

  The processor 21, the memory 22, the hard disk device 23, the display device 24, the input device 25, the optical drive device 26, and the network interface 28 illustrated in FIG. 12 are connected to one another via a bus. Further, the processor 21, the peripheral device interface 29, and the robot interface 30 are also connected to each other via the bus described above. The sensor unit SD is connected to the peripheral device interface 29 illustrated in FIG. 12, and the processor 21 acquires the measurement result obtained by the sensor unit SD via the peripheral device interface 29. The robot interface 30 shown in FIG. 12 is connected to the bear-type robot CR, and the processor 21 exchanges information with the robot CR via the robot interface 30.

  The optical drive device 26 described above can be mounted with a removable disk 27 such as an optical disk, and reads and records information recorded on the mounted removable disk 27. Moreover, the input device 25 illustrated in FIG. 12 is, for example, a keyboard or a mouse. An operator of the interaction device 10 can input an instruction for specifying an object to be achieved by presenting behavior to the user U, for example, by operating the input device 25. Note that the operator of the interaction device 10 is, for example, a person who requests the above-described user U watching service. The operator of the interaction device 10 may live in a place away from the place where the interaction device 10 is installed. An operator who is away from the interaction device 10 may input an instruction to the interaction device 10 via the network NW and the network interface 28, for example.

  The memory 22 illustrated in FIG. 12 stores an application program for executing an interaction process for the processor 21 to interact with the user U as described above, together with the operating system of the computer device 20. Note that the application program for executing the interaction process and the information for constructing the database 13 shown in FIG. 7 can be recorded and distributed on, for example, a removable disk 27 such as an optical disk. Then, an application program for executing the interaction process may be stored in the memory 22 and the hard disk device 23 by loading the removable disk 27 in the optical drive device 26 and performing a reading process. The database 13 may be generated in the hard disk device 23 based on the information for constructing the database 13 described above together with the installation work of the application program for executing the interaction process. The application program for executing the interaction process and the information for constructing the database 13 can also be acquired via the network interface 28.

  The application program for the interaction process includes an application program for causing the processor 21 to execute a process for detecting a situation related to the interaction between the robot CR and the user U.

  The processor 21 implements the function of the detection unit 12 by periodically executing an application program for processing for detecting the situation over a period in which the user U is acted on via the robot CR. May be.

  FIG. 13 shows an example of a flowchart of processing for detecting a situation. Each process of step 201 to step 205 illustrated in FIG. 13 can be realized by the processor 21 executing an application program for a process for detecting a situation.

  First, the processor 21 collects measurement values acquired by various sensors provided in the sensor unit SD and the robot CR shown in FIG. 12 (step 201). In step 201, the processor 21 collects various measurement values related to the environment such as brightness and temperature obtained by the sensor unit SD via the peripheral device interface 29. Further, the processor 21 collects measurement values indicating the reaction of the user U obtained by a camera, a microphone, or the like provided in the robot CR via the robot interface 30.

  Next, the processor 21 calculates each index value indicating the state of the user U based on the measured value indicating the reaction of the user U obtained in the process of step 201 (step 202). As described above, in this step 202, the processor 21 can use the technique of the above-mentioned Japanese Patent Application No. 2011-243573.

  Next, the processor 21 stores the interaction information including various measured values related to the environment collected in step 201 and each index value calculated in step 202 in the status table 123 provided in the hard disk device 23 or the like at the current time. Store correspondingly (step 203).

  Next, after waiting for the next detection timing in step 204, the processor 21 determines whether or not an instruction to stop the interaction processing is input via the input device 25 shown in FIG. 12, for example (step 205). ). When the instruction to stop the interaction process is not input, the processor 21 returns to the process of step 201 according to the negative determination route of step 205 and repeats the processes of step 201 to step 205.

  In this way, the processor 21 detects the situation related to the interaction between the robot CR and the user U every predetermined period, and the situation table in which the interaction information obtained by this detection processing is provided in the hard disk device 23 or the like. 123 can be held. In addition, in the process of detecting the above-described situation, the processor 21 executes the process of step 202, thereby realizing the function of the analysis unit 122 illustrated in FIG. Thereby, the processor 21 can hold the interaction information including the index values indicating the state of the user U in the situation table 123 described above.

  Note that, for example, when an instruction to end the interaction process is input after the confirmation that the user U has gone to bed, the processor 21 changes the situation according to the affirmative determination route of step 205 described above. The detection process is terminated.

  Further, the application program for the interaction process includes an application program for causing the processor 21 to execute a process for controlling the behavior presented to the user U by the robot CR.

  FIG. 14 shows an example of a flowchart of processing for controlling behavior. Each process of step 301 to step 306 shown in FIG. 14 can be realized by the processor 21 executing each application program included in the application program for the process for controlling the behavior.

  The application program for the process for controlling the behavior includes an application program for causing the processor 21 to perform a process for selecting the behavior from the database 13 in accordance with an instruction for specifying the purpose of the interaction. In the flowchart shown in FIG. 14, the processor 21 waits for an instruction to specify the purpose of interaction in step 301, and then executes the application program in step 302. In step 302, a behavior that can achieve the purpose specified by the instruction received in step 301 is selected from the database 13 provided in the hard disk device 23. That is, the execution of the processing of step 302 by the processor 21 is an example of a technique for realizing the search unit 14 and the selection unit 15 illustrated in FIG. 1 by the cooperation of the processor 21, the memory 22, and the hard disk device 23. is there.

  Further, the application program for the process for controlling the behavior is an application program for causing the processor 21 to execute the process of presenting the selected behavior to the user U using the function provided in the robot CR. Contains. In the flowchart shown in FIG. 14, the processor 21 executes this application program in step 303. Then, the processing selected in step 302 causes the output device 11 realized by the robot CR to execute the behavior selected in step 302. That is, the execution of the process of step 303 by the processor 21 is an example of a technique for realizing the control unit 16 illustrated in FIG. 1 by the cooperation of the processor 21, the memory 22, and the hard disk device 23.

  Furthermore, the application program for the process for controlling the behavior may include an application program for causing the processor 21 to execute a process for updating the expected degree for the purpose of the outputted behavior based on the detected change in the situation. Good. If it is determined in step 304 of the flowchart shown in FIG. 14 that the purpose is specified, the processor 21 executes the application program in step 305. Then, by the processing in step 305, the processor 21 updates the expected degree of the purpose of the behavior executed in the processing in step 303 based on the interaction information in the status table 123 provided in the memory 22 or the hard disk device 23. To do. That is, the execution of the processing of step 305 by the processor 21 is an example of a technique for realizing the updating unit 17 illustrated in FIG. 7 by the cooperation of the processor 21, the memory 22, and the hard disk device 23.

  Further, in step 306 of the flowchart shown in FIG. 14, the processor 21 determines whether or not an instruction to stop the interaction processing is input through the input device 25 shown in FIG. 12, for example. When the instruction to stop the interaction process is not input, the processor 21 returns to the process of step 301 according to the negative determination route of step 306 and repeats the processes of step 301 to step 306. In this way, regardless of the presence / absence of an instruction for specifying a purpose, the robot CR can be continuously acted on the user U. Then, for example, when an instruction to end the interaction process is input, triggered by the confirmation that the user U has gone to bed, etc., the processor 21 ends the process according to the affirmative determination route of step 306.

  Next, the process of selecting the behavior to be executed by the robot CR will be described using the flowchart of FIG.

  FIG. 15 shows an example of a flowchart of processing for selecting a behavior. Each process of step 311 to step 317 illustrated in FIG. 15 is an example of a process included in an application program for selecting a behavior. In addition, each processing of step 311 to step 317 is executed by the processor 21.

  First, the processor 21 obtains the latest interaction information by referring to the situation table 123 provided in the memory 22 or the hard disk device 23 (step 311).

  Next, the processor 21 extracts the behavior satisfying the trigger condition by collating the latest interaction information acquired in step 311 with the trigger condition included in each behavior definition information registered in the database 13. (Step 312). As described above, the interaction information held in the situation table 123 includes each index value indicating the state of the user U. Therefore, the processor 21 can select the behavior definition information to be extracted based on the trigger condition including the condition related to the state of the user U by executing the process of step 312.

  Next, the processor 21 determines whether or not an instruction for specifying a purpose is input in the process of step 301 shown in FIG. 14 (step 313).

  When an instruction to specify the purpose of the interaction is input by the operator of the interaction device 10, the processor 21 executes the processing of step 314 to step 316 according to the affirmative determination route of step 313.

  In step 314, the processor 21 narrows down the behavior candidates from the behavior extracted in step 312 based on whether or not the expectation level corresponding to the designated purpose is set in the behavior definition information. When a plurality of purposes are specified, the processor 21 includes the behavior corresponding to the behavior definition information in which the degree of expectation for at least one of these purposes is set as the behavior candidate. In this way, the processor 21 can perform the function of the search unit 14 shown in FIG. 1 by executing the processing of step 314 following step 312 described above.

  Next, the processor 21 selects individual behavior candidates using the above-described equation (1) based on the expectation degree set in correspondence with the designated purpose in the behavior definition information corresponding to each behavior candidate. The probability of being played is calculated (step 315). The processor 21 can perform the function of the calculation unit 151 illustrated in FIG. 4 by executing the processing of step 315. Note that, in the case where instructions for specifying a plurality of purposes are input in step 301 shown in FIG. 14, the processor 21 uses these equations (2) in the course of the processing in step 315 to determine these purposes. A composite expectation is calculated based on the expectation corresponding to. Further, in this case, the processor 21 calculates the probability represented by the equation (4) using this synthesis expectation.

  Then, the processor 21 determines the behavior to be output to the robot CR by selecting one of them according to the probability calculated for each behavior (step 316). In the process of step 316, the processor 21 realizes the function of the determination unit 152 shown in FIG. 4, and determines a behavior to be selected from a plurality of behavior candidates according to the probability reflecting the expectation. Thereby, it is possible to maintain the diversity of behaviors output by the robot CR while increasing the possibility of selecting behaviors that are highly likely to achieve the purpose.

  As described above, in the affirmative determination route of step 313, the processor 21 can select the behavior to be output to the robot CR in order to achieve the specified purpose by executing the processing of step 314 to step 316. .

  On the other hand, in the negative determination route in step 313, the processor 21 refers to the priority corresponding to each behavior extracted in the processing in step 312 and selects, for example, the behavior having the highest priority as the behavior to be output to the robot CR. (Step 317). As described with reference to FIG. 2, the behavior definition information corresponding to each behavior includes a priority according to the property of the behavior. In the process of step 317, the processor 21 responds to the action from the user U rather than the behavior that is continuously executed, for example, the breathing motion, by selecting the behavior according to the priority. Priority is given to the behavior.

  As described above, in the negative determination route of step 313, the processor 21 can select the behavior to be executed by the robot CR in order to make the robot CR look like a living thing by executing the processing of step 317.

  Based on the behavior definition information corresponding to the behavior selected in this way, the processor 21 causes the robot CR to execute the behavior selected in step 303 shown in FIG. In step 303, the processor 21 passes the behavior content included in the behavior definition information to the robot CR via the robot interface 30 shown in FIG. 12, thereby causing the robot CR to execute the behavior based on the behavior content. In step 303, for example, the processor 21 sets the time before and after the execution of the behavior to the time when the behavior content is passed to the robot CR and the time when the notification of the completion of the behavior output is received from the robot CR. May be stored in the memory 22 or the like.

  When the behavior executed by the robot CR in step 303 is the behavior for achieving the specified purpose, the processor 21 proceeds to the affirmative determination route in step 304 and updates the expectation in step 305. Execute the process.

  FIG. 16 shows an example of a flowchart of a process for updating the expectation degree. Each process of step 321 to step 324 illustrated in FIG. 16 is an example of a process included in the application program for the process of updating the expectation degree. In addition, each processing of step 321 to step 324 is executed by the processor 21.

  First, the processor 21 refers to the situation table 123 described above, and acquires the interaction information held corresponding to the time before and after the execution of the behavior (step 321). The time before and after the execution of the behavior can be obtained by, for example, storing it in the memory 22 or the like in the process of step 303 described above. In addition, when the processor 21 causes the robot CR to execute the behavior for achieving the specified purpose in the processing of step 303, the processor 21 executes the behavior by referring to the status table 123 prior to the transfer of the behavior contents. The previous interaction information may be acquired.

  Next, the processor 21 compares the index value corresponding to the designated purpose included in the interaction information before and after the execution of the behavior, thereby obtaining a change in the degree of achievement of the goal indicated by the change δ in the index value. Calculate (step 322).

  Next, the processor 21 calculates a new expectation level using the above-described formula (5) based on the change δ of the index value calculated in the process of step 322 and the expectation level corresponding to the index value. Calculate (step 323). As described above, the method in which the processor 21 calculates a new expectation in step 323 is not limited to the method using the equation (5).

  Then, the processor 21 performs a process of rewriting the expectation included in the behavior definition information for the behavior definition information corresponding to the behavior executed in step 303 using the new expectation calculated in step 323 (step 324).

  The processor 21 executes the processing from step 321 to step 324 described above, thereby realizing the function of the updating unit 17 illustrated in FIG. Then, by performing the expectation level update process as described above, customization according to the individuality of the user U can be applied to the interaction device 10 in the course of the interaction between the robot CR and the user U. This is as described above.

  Each of the embodiments described above is an example of an embodiment of the interaction device disclosed herein. For example, the individual technical elements described with reference to FIGS. 4, 7, and 11 are combined in different combinations. By applying this, it is possible to make another embodiment.

  For example, when the behavior definition information in which the trigger condition including the condition related to the state of the user U is set is not registered in the database 13, the analysis unit 122 provided in the detection unit 12 does not always have the user U regularly. It is not necessary to calculate an index value related to the state of. Then, for example, in response to a request from the updating unit 17 or the control unit 16, an index value related to the state of the user U is calculated at timings before and after execution of the behavior, and the obtained index value is passed to the updating unit 17. May be.

Regarding the above description, the following items are further disclosed.
(Supplementary note 1) an output device that outputs a plurality of behaviors including at least one of voice and action;
A detection unit for detecting a situation related to an interaction between the person who is acted on by the output device and the output device;
Corresponding to each behavior, a behavior definition including a trigger condition indicating a condition for outputting the behavior to the output device, and type information indicating a type of state change of the person that can be extracted by the output of the behavior A database for registering information;
In response to an instruction designating at least one occurrence of the plurality of types of state changes as a purpose of acting on the person, behavior definition information in which the trigger condition is satisfied by a situation detected by the detection unit, And the search part which searches the behavior definition information including the type information indicating the state change of the specified type from the database as a behavior candidate,
A selection unit for selecting one of the behavior candidates;
An interaction device comprising: a control unit that causes the output device to execute a behavior corresponding to the behavior definition information selected by the selection unit.
(Appendix 2) In the interaction device described in Appendix 1,
The behavior definition information corresponding to each behavior includes a degree of expectation indicating a high possibility of causing a state change of each type indicated by the type information,
The selection unit includes:
In the behavior definition information corresponding to each of the behavior candidates, one behavior is selected from the behavior candidates according to the probability having a correlation with the high degree of expectation for the type of state change specified by the instruction. A featured interaction device.
(Appendix 3) In the interaction device described in Appendix 2,
The interaction device, wherein the selection unit calculates a probability that the behavior definition information is selected so as to be proportional to the expectation included in the behavior definition information.
(Appendix 4) In the interaction device described in Appendix 2,
The selection unit includes:
When a plurality of purposes are specified in the instruction, a totaling unit that totals the expectation degree for the plurality of purposes included in each of the behavior definition information,
An interaction device, characterized in that the probability that each of the behavior definition information is selected is calculated so as to be proportional to a total result obtained by the total unit.
(Appendix 5) In the interaction device according to any one of Appendix 1 to Appendix 4,
The detector is
An analysis unit that obtains each of a plurality of types of index values representing the state of the person by analyzing measurement values indicating the reaction of the person obtained by observing the person U;
The plurality of types of index values obtained by the analysis unit are output as a part of information indicating a detected situation,
The behavior definition information corresponding to each behavior includes, as a part of the trigger condition, information indicating a value range of the index value for at least one of the plurality of index values.
(Appendix 6) In the interaction device described in Appendix 5,
Based on the change of the index value obtained by the analysis unit before and after the output of the behavior, the behavior definition information used for the output of the behavior is about the type of state change designated as the purpose of the action in the instruction. An interaction device comprising: an update unit that updates an expectation degree.
(Supplementary note 7) Detecting a situation relating to an interaction between the output device and the person subjected to the action performed by the output device outputting behavior including at least one of sound and action,
Corresponding to each behavior, a behavior definition including a trigger condition indicating a condition for outputting the behavior to the output device, and type information indicating a type of state change of the person that can be extracted by the output of the behavior From the database to register information,
In response to an instruction designating at least one occurrence of the plurality of types of state changes as a purpose of acting on the person, behavior definition information in which the trigger condition is satisfied by a situation detected by the detection unit, And the behavior definition information including the type information indicating the state change of the specified type is searched from the database as a behavior candidate,
Select one of the behavior candidates,
An interaction control program for causing a process corresponding to the selected behavior definition information to be executed by the output device.
(Appendix 8) In the interaction control program described in Appendix 7,
An interaction control for executing a process of updating an expectation degree related to the designated purpose included in the behavior definition information corresponding to the behavior based on a change in a situation detected according to the output of the behavior; program.

DESCRIPTION OF SYMBOLS 10 ... Interaction apparatus; 11 ... Output device; 12 ... Detection part; 13 ... Database; 14 ... Selection part; 15 ... Control part; 17 ... Update part; 121 ... Collection part; Calculating unit; 152 ... Determination unit; 153 ... Counting unit; 20 ... Computer device; 21 ... Processor; 22 ... Memory; 23 ... Hard disk device (HDD); 24 ... Display device; 25 ... Input device; 27 ... Removable disk; 28 ... Network interface; 29 ... Peripheral device interface; 30 ... Robot interface; 31 ... Drive processing unit; 32 ... Motor; 33 ... Speaker; 34 ... Observation unit; Robot; U ... User; SD ... Sensor part; NW ... Network

Claims (5)

An output device that outputs a plurality of behaviors including at least one of sound and action;
A detection unit for detecting a situation related to an interaction between the person who is acted on by the output device and the output device;
Corresponding to each behavior, a behavior definition including a trigger condition indicating a condition for outputting the behavior to the output device, and type information indicating a type of state change of the person that can be extracted by the output of the behavior A database for registering information;
In response to an instruction designating at least one occurrence of the plurality of types of state changes as a purpose of acting on the person, behavior definition information in which the trigger condition is satisfied by a situation detected by the detection unit, And the search part which searches the behavior definition information including the type information indicating the state change of the specified type from the database as a behavior candidate,
A selection unit for selecting one of the behavior candidates;
An interaction device comprising: a control unit that causes the output device to execute a behavior corresponding to the behavior definition information selected by the selection unit. The interaction device according to claim 1,
The behavior definition information corresponding to each behavior includes a degree of expectation indicating a high possibility of causing a state change of each type indicated by the type information,
The selection unit includes:
In the behavior definition information corresponding to each of the behavior candidates, one behavior is selected from the behavior candidates according to the probability having a correlation with the high degree of expectation for the type of state change specified by the instruction. A featured interaction device. The interaction device according to claim 1 or 2,
The detector is
By analyzing a measurement value indicating the reaction of the person obtained by observing the person, an analysis unit for obtaining each of a plurality of types of index values representing the state of the person,
The plurality of types of index values obtained by the analysis unit are output as a part of information indicating a detected situation,
The behavior definition information corresponding to each behavior includes, as a part of the trigger condition, information indicating a value range of the index value for at least one of the plurality of index values. The interaction device according to claim 3,
Based on the change of the index value obtained by the analysis unit before and after the output of the behavior, the behavior definition information used for the output of the behavior is about the type of state change designated as the purpose of the action in the instruction. An interaction device comprising: an update unit that updates an expectation degree. Detecting the situation regarding the interaction between the person who receives the action performed by the output device outputting the behavior including at least one of voice and action, and the output device;
Corresponding to each behavior that can be output by the output device, a trigger condition that indicates a condition for causing the output device to output the behavior, and a type that indicates the type of state change of the person that can be extracted by the output of the behavior From the database that registers the behavior definition information including information,
In response to an instruction designating at least one occurrence of the plurality of types of state changes as a purpose of acting on the person, behavior definition information in which the trigger condition is satisfied by a situation detected by the detection unit, And the behavior definition information including the type information indicating the state change of the specified type is searched,
Select one of the behavior definition information obtained as a search result,
An interaction control program for causing a computer to execute a process for causing the output device to execute a behavior corresponding to the selected behavior definition information.

Images