ABSTRACT
Chapter 5 provides a multimodal teaching-by-demonstration system, which transfers themultimodal information adaptation ability from a human tutor to the robot. Based on the limb surface electromyography signals obtained from the demonstration phase, the human tutor's stiffness is estimated. And the force profiles in Cartesian space are collected from a force/torque sensor mounted between the robot endpoint and the tool. Then, the hidden semi-Markov model is used to encode the multiple signals in a unified manner and encode the correlations between position and the other three control variables (i.e., velocity, stiffness, and force) separately. The expected control variables are generated by the Gaussian mixture regression based on the estimated parameters of the HSMM model. To reproduce the task, the learned variables are further mapped into an impedance controller in the joint space through inverse kinematics. Finally, the effectiveness of the approach is proved by experiments.
