Primitive communication of humanoid robot with human via hierarchical mimesis model on the proto symbol space

Abstract

Communication is defined as a process of information exchange between individuals through some common systems such as symbols, signs, languages, or gestures. This communication skill is important for social robots which interact with humans. Especially primitive communication consists of interaction patterns between 2 humans. In this paper we propose a mimetic communication model, which is a hierarchical model based on mimesis model that integrates motion recognition with motion generation by hidden Markov models (HMMs). HMMs in the lower layer abstract the motion patterns and HMMs in the upper layer represent the interaction patterns through the lower proto symbols. We also demonstrate the validity of this model through several experiments on a humanoid robot