A human-demonstration based approach to the recognition of process state transitions in the insertion of a deformable tube is presented. First, human demonstration of the insertion of a deformable hose into a rigid plug is measured by use of a position sensor and a force sensor. Secondly, the measurements are analyzed with regard to the process states during the insertion. Human recognition laws to detect process state transitions from force sensation are then extracted through the analysis of human demonstration. Finally, human motion during the insertion is transplanted to a mechanical manipulator to examine whether the human recognition laws are useful or not for the insertion performed by mechanical manipulators.

One approach to the transplantation of human skilful motion to a manipulator is presented. First, human motion during the insertion of a deformable hose into a plug is measured by use of a position sensor and a force sensor. The measurements are then analyzed with regard to the contact between the hose and the plug. It is found that an alternating motion during the process is effective on the successful insertion of a deformable hose into a plug. Next, human motion is embedded in a manipulator program directly. It turns out that the manipulator can perform the insertion successfully as far as the positional error of the hose is involved in a certain range. Finally, it is shown that refining human motion helps a manipulator perform the operation more quickly.