Abstract:

This dissertation aims to develop an intelligent safe human–robot interaction (sHRI) system and apply it to an active–passive variable stiffness elastic actuator (APVSEA). A Bond graph is used to construct the system model, model matching controller (MMC), and robust fault detection and isolation system. To import the human factor into the sHRI system, Kinect was adopted to detect points on the human skeleton. Human joint positions and their velocities in space are used to dynamically adjust the actuator stiffness for sHRI.

MMC can force the output response of a plant to that of a reference. In this study, a complete MMC design flowchart is proposed. Moreover, an MMC was implemented in the APVSEA system. The MMC is used to change a plant stiffness so that if a human collides with a robot, the human will not sustain injuries.

If any key system components break or fail, the entire system may destabilize or become divergent. Thus, this study develops a robust fault detection and isolation (RFDI) system for effectively detecting key component faults. When a fault in the system is detected, the RFDI system is switched to a suitable control system to guarantee human safety.

In summary, this study proposes an intelligent sHRI system that can vary the stiffness of a plant and detect fault components. Furthermore, by importing the human factor, the sHRI system becomes even more smart.