Modeling of two-fingered pivoting skill based on CPG

Two-fingered pivoting is a typical example of human dexterous manipulation interacting adaptively with the environment. In this paper, we construct a skill model of two-fingered pivoting manipulation based on CPG (central pattern generator) with five neurons. The CPG drives a virtual hand, which consists of a wrist, a palm, a thumb and an index finger, rhythmically to achieve robust pivoting through the interaction with the environment. Its adaptability is demonstrated in dynamic simulation of box pivoting. The virtual hand successfully adapts to environmental changes such as sudden decrease of the friction coefficient of the floor.