The authors develop a method of modeling a human manipulativeskill using human linguistic knowledge about the task. A globalnonlinear structure of human control behavior is constructed based onthe linguistic information, and all functionalities used by thelinguistic structure are identified from human demonstration data.Mapping between sensor space and human mental space for input signals isdiscussed to elucidate human skills. Techniques for selectingsignificant features extracted from sensor signals and reducing thedimension of the sensor space are developed. The techniques were appliedto a direct-drive deburring robot to verify the feasibility of themethod
展开▼