Perception of sound, or audition, for robots is significantly important to communicate with human. In our lab, a fundamental robot action schemes for robots with auditory sensors are investigated taking dynamics of robot itself into account.
As a fundamental motion primitive for sound source localization by an auditory robot, a direct feedback of the acoustic cues to control the robot head to point the sound source is developed.
Such robot motions and changes in the environment may change the transfer characteristics of acoustic signals from the sound source to the robot, which may deteriorate the localization performance. In order to overcome this issue, it is proposed to model the effect by the change of the characteristics as the uncertainty which is taken into account by the sound localization algorithms. As another approach to reduce the effect of the uncertainty, active pinnae that was inspired from the fact some animals are able to change the shape of their pinnae to improve the sound localization performance, has been also developed.