Visually-Guided Control for Safe Human-Robot Interaction

Since past decades, it was predicted that robots will eventually enter diverse daily environments and closely interact with humans. Recent technological advances in robotics enabled robots possible to realize systems of high integration, advanced perceptional capabilities and enhanced power to cross this barrier. Various companies have developed their first commercial service robotic products in recent years. However, the increasing effort has made it necessary to address the fundamental requirements of service robots in domestic environment. They range from useful applications, and robust operation in the unstructured environment, to the human-robot interaction experience. This dissertation is devoted to complying the necessary requirements with applications to the four topics: the robotic illuminating assistant, safe human-robot interaction, safe humanoid manipulation, and imitating human-like behavior by motor resonance. In response to the requirement of useful household applications for service robotics, we suggest an idea that robots should support us through enhancing the way we interact with household appliances, instead of replacing us at certain chores. Therefore, we proposed to use a robotic manipulator to serve as the household appliance, e.g. robotic illuminating assistant, for interactively illuminating support by using the eye-in-hand visual servoing. For robust operation in the unstructured environment, it intrinsically needs to consider the safety in human-robot interaction and the environmental perception. For assuring human-robot-environment safety, the virtual impedance control and the safe humanoid manipulation are proposed. The virtual impedance control is designed for the robust, smooth, and consistent collision avoidance that the robot can avoid all the possible collisions robustly while trying to accomplish the original task. By monitoring the workspace, the robots can respond to the human nearby safely and compliantly. In addition, as humanoids operating in unstructured environments, it is necessary to simultaneously consider not only robust task execution but also the safety issues, such as colliding with objects in the environment, self-collision, and so on. In this case, we proposed a multi-priority reactive control framework based on the virtual impedance control approach for safe humanoid manipulation. Finally, to achieve better human-robot interaction experience, the robot needs to behave predictably compared to human beings beside safety issues. We teach the humanoid robot imitating human-like behavior by motor resonance. The robot can perform sign languages by observing human demonstration for the better experience of interacting with deaf people.