Generation of reactive stepping motion for a humanoid by dynamically stable mixture of pre-designed motions

Introduces and describes a method which enables humanoid robots to make the stable reactive motion of stepping to an arbitrary position. Taking dynamical stability into consideration, we have realized such stepping motion in two stages. The first stage is to construct kernel motions using a genetic algorithm in a dynamics simulation environment. The second stage is to make a real-time mixture of the kernel motions without breaking dynamical stability. Mixed motions enable the robots to step to desired positions determined by sensory information. The validity of this method has been verified in an experiment in which a bipedal humanoid robot approaches a target object by making stepping motion based on visual information.