Behavior control device, behavior control method, and program
A social ability generation device includes a perception unit that acquires person information on a person, extracts feature information on the person from the acquired person information, perceives an action that occurs between a communication device performing communication and the person, and perceives an action that occurs between people, a learning unit that multimodally learns an emotional interaction of the person using the extracted feature information on the person, and an operation generation unit that generates a behavior on the basis of the learned emotional interaction information of the person.
1. A social ability generation device comprising:
a perception unit configured to acquire person information on a person, extract feature information on the person from the acquired person information, perceive an action that occurs between a communication device performing communication and the person, and perceive an action that occurs between people;
a learning unit configured to multimodally learn an emotional interaction of the person using the extracted feature information on the person; and
an operation generation unit configured to generate a behavior on the basis of the learned emotional interaction information of the person,
wherein the learning unit
performs learning using an implicit reward and an explicit reward,
generates an implicit reaction system using the implicit reward and a social model,
generates an explicit reaction system using the explicit reward according to a social structure or social norm while evaluating the action of the communication device, and
performs an output incorporating at least one of the social model and a social structure concept by using the implicit reaction system and the explicit reaction system at a time of operating,
the implicit reward is a multimodally learned reward using the feature information on the person, and
the explicit reward is a reward based on a result of evaluating a behavior of the communication device with respect to the person generated by the operation generation unit.
2. The social ability generation device according to claim 1 , further comprising:
a sound pickup unit configured to pick up an acoustic signal; and
an imaging unit configured to capture an image including a user,
wherein the perception unit performs speech recognition processing on the picked-up acoustic signal to extract feature information on voice, and performs image processing on the captured image to extract feature information on a human behavior included in the image,
the feature information on the person includes the feature information on a voice and the feature information on the human behavior,
the feature information on a voice is at least one of an audio signal, information on volume of sound, information on intonation of the sound, and a meaning of utterance, and
the feature information regarding a human behavior is at least one of facial expression information of the person, information on a gesture performed by the person,
head posture information of the person, face direction information of the person, line-of-sight information of the person, and a distance between people.
3. The social ability generation device according to claim 1 , wherein the learning unit performs learning using social norms, social components, psychological knowledge, and humanistic knowledge.
4. A social ability generation method comprising:
acquiring, by a perception unit, person information on a person, extracting feature information on the person from the acquired person information, perceiving an action that occurs between a communication device performing communication and the person, and perceiving an action that occurs between people;
multimodally learning, by a learning unit, an emotional interaction of the person using the extracted feature information on the person;
generating, by an operation generation unit, a behavior on the basis of the learned emotional interaction information of the person;
performing, by the learning unit, learning using an implicit reward and an explicit reward;
generating, by the learning unit, an implicit reaction system using the implicit reward and a social model,
generating, by the learning unit, an explicit reaction system using the explicit reward according to a social structure or social norm while evaluating the action of the communication device, and
performing, by the learning unit, an output incorporating at least one of the social model and a social structure concept by using the implicit reaction system and the explicit reaction system at a time of operating,
the implicit reward is a multimodally learned reward using the feature information on the person, and
the explicit reward is a reward based on a result of evaluating a behavior of the communication device with respect to the person generated by the operation generation unit.
5. A communication robot comprising:
a perception unit configured to acquire person information on a person, extract feature information on the person from the acquired person information, perceive an action that occurs between a communication device performing communication and the person, and perceive an action that occurs between people;
a learning unit configured to multimodally learn an emotional interaction of the person using the extracted feature information on the person; and
an operation generation unit configured to generate a behavior on the basis of the learned emotional interaction information of the person,
wherein the learning unit
performs learning using an implicit reward and an explicit reward,
generates an implicit reaction system using the implicit reward and a social model,
generates an explicit reaction system using the explicit reward according to a social structure or social norm while evaluating the action of the communication device, and
performs an output incorporating at least one of the social model and a social structure concept by using the implicit reaction system and the explicit reaction system at a time of operating,
the implicit reward is a multimodally learned reward using the feature information on the person, and
the explicit reward is a reward based on a result of evaluating a behavior of the communication device with respect to the person generated by the operation generation unit.
6. The communication robot according to claim 5 , further comprising:
a display unit,
wherein the operation generation unit generates an image that maintains a relationship with the person as a good relationship by causing the person to perform a behavior for maximizing positive emotion, and displays the generated image on the display unit.