This study aims to develop methods to increase information available to a haptic device about a human operator during physical human-robot interaction. Physical contact between a robot and human operator establishes a coupled system with stability and performance characteristics partly dependent on interaction port impedance behavior. Operator impedance is estimated in this research based on changes in arm muscle activity, recorded through electromyographic (EMG) signals. A switching impedance controller employing a Support Vector Machine (SVM) classifier for operator state estimation is used in an interaction system with a one degree-of-freedom haptic device. Results from performance (e.g. speed, accuracy) trials investigating a stochastic approach to position control are presented in comparison to other standard approaches. This research serves a basis for the exploration of advanced control tools and ultimately developing a novel human-robot interface. Applications for such research include interaction with robot co-workers (e.g. power-assisting devices) in industrial settings.
Attention-Oriented Action Recognition for Real- Time Human-Robot Interaction
