Development of character robots for human-robot mutual communication

This paper describes a robot-head system as a communication device for human robot interaction. Most robotic systems with natural modalities have facial expression functions, since facial expressiveness is regarded as a key component to develop personal attachment along with prosodic expressiveness. Most of the conventional facial robots has adopted Ekman's FACS. However, due to some mechanical constraints, there are some limitations to adopt FACS model completely in facial robots. In the first part of this paper, we introduce a character robot - CRF2 that have richer facial expressions than the first prototype character robot -CRFl since the facial design of CRF2 implements eyelids' mechanism and it is based on the 3D deformation model. However due to the fact that the recognition rate of some expressions such as disgust and fear was not improved very much, we concluded that character robot needs other communication channels such as voice to convey robot's emotional states properly to the user in addition to renovating mechanical facial part. As a result we developed a renovated character robot -CRF3. In addition to facial expressiveness, CRF3 is implemented speech synthesis and neck motions. Further CRF3 has visual, auditory and tactile sensors and expanded configuration that can connect additional sensor or actuator modules to the system. To evaluate the applications of CRF, we applied CRF3 to home environments as a home network manager which selects a task based on robot mood.