Abstract
Human-centered design and control are vital for the human-machine system. Especially, the estimation of human's intention will become a key technology for the cooperation with human and the physical assistance by robotic machine. Surely, we are regulating well the mechanical impedance (inertia, viscosity and stiffness) of our arm in accordance with the task. Therefore, it is supposed that the adjustment of impedance reflects the operational intention. In this study, the positioning task with different accuracy is treated. And, the hand-viscosity is estimated on-line during positioning by the neural network whose input is the electromyogram (EMG). Using this information, the dynamics of machine is controlled so as to generate adequately assistant motion. In the experiments, the effectiveness of the proposed operational assistance system is confirmed from the viewpoint of maneuverabilities and workload.
