A novel cognitive architecture for a human-like virtual player in the mirror game

Abstract

The so called mirror game, which in its simplest formulation involves two people mirroring each other's hand's movement, provides a paradigm to study social interaction. However, a customized virtual player can replace either of the two human participants and hopefully help with the rehabilitation of patients suffering from social disorders by regulating its kinematics. In this paper we investigate the coordination movement between an avatar (virtual player) and a human player in the mentioned game. A novel cognitive architecture is proposed to drive the motion of the virtual player so that it generates a human-like trajectory in two different experimental models. In order to achieve this objective, the Haken-Kelso-Bunz (HKB) equation is adopted to describe the social motor coordination between the virtual and the human player. In addition, both an adaptive algorithm for the coupling parameters in the HKB equation and a feedback controller are developed in order to guarantee human features for the virtual player in its kinematics. Finally, extensive experiments are conducted to validate the approach described above.