A new robot motion authoring method using HTM

This paper presents a user-friendly authoring method for humanoid robots using HTM (hierarchical temporal memory), a machine-learning model. The proposed method is an intuitive scheme which allows users (especially children and the elderly who are interested in robots but do not have enough knowledge) to author (create and edit) various motions for the humanoid robot. The new method can derive in real time various motions from the robot postures that users create by turning a dial based on the learning and inference procedures of HTM. Simulation results for authoring a humanoid robot with 26 degrees of freedom (DOF) is given to show the effectiveness of the proposed method.