A sentential cognitive system of robots for conversational human-robot interaction

Abstract

For interactions with humans, robots need the capability of describing the events they encountered for verbal communication. In this paper, a sentential cognitive system (SCS) using sentences as a media to represent cognized events for conversational human–robot interaction (HRI) is presented. The SCS comprising multiple modules such as perception and behavior, reasoning, and memory interprets modular events to sentential form from acquired cognitive information and a cognitive grammar linking it to sentences and vice versa. The sentences are stored in a sentential memory for being retrieved by reasoning procedures with the auxiliary memories, an object descriptor and a motion descriptor, afterward for conversational HRI. In the experiment, the proposed SCS is implemented in a robot and tested with scenarios of spatiotemporal conversation about the handling of objects. The result shows the feasibility and efficiency of the proposed SCS for conversational HRI.