Human-Inspired Audiovisual Inducement of Whole-Body Responses

For service humanoid robots, it is crucial to design human-robot interaction behaviors that avoid eliciting negative responses from users. Drawing inspiration from the natural human response of turning attention when hearing one's own name, this paper introduces a system that enables users to effortlessly capture a robot's attention, fostering a more natural and intuitive interaction. The proposed system consists of an integrated audio-visual perception system and a sophisticated whole-body controller, effectively merging audio and vision data to stimulate comprehensive and coordinated robot motion during human robot interaction scenario. The system has been successfully implemented and tested on the service humanoid robot GARMI. The resultant whole-body motion emulates the natural human process of localizing sound and visual cues, enabling the robot to track and follow the user who initiated the interaction. Moreover, the system's capabilities have been extended to include an object-reaching use-case, effectively demonstrating the versatility of the whole-body controller. To enhance user-friendliness, we have incorporated a natural language command interface, allowing users to effortlessly control the activation of the proposed system while serving as an attention-switching mechanism. Moreover, the robot employs an intuitive audio-visual feedback mechanism, offering transparency about its current state to users. Lastly, the system's performance has been rigorously evaluated through a series of experiments, confirming its effectiveness and reliability.