Human Robot Interactions

Physical Human-Robot Interactions


Human Intent Estimation  


 This study aims to maximize the free space transparency in physical human-robot interaction (pHRi). A primary method to achieve this objective is to mathematically formulate the motion intended by the human and input it into the robot controller. To this end, human dynamics analysis and physical interaction analysis were conducted. Based on these analyses, an input compensator was designed for the robot to follow the human's intentions.


Human Impedance Compensation


 A compensator for physical human-robot interaction has been developed using the SPSA and AMSGrad optimization algorithms. The overall algorithm is based on admittance control. Unlike the traditional method, which estimated the impedance of a human hand through simple experiments, this improved method employs optimization for impedance compensation. Experiments were conducted using a 2-axis hydraulic robot.






Force Control with Learning from Demonstration


Learning from Demonstration


 Learning from demonstration is a control technique that mimics human task performance. The work trajectory and external force required for the work are input through physical interaction with the collaborative robot. Tasks such as grinding and writing are difficult tasks that involve external forces acting on the work object and require human know-how and senses. Research is being conducted to input these tasks so that robots can directly imitate them and control them repeatedly.



Multi-Robot Cooperative Manipulation


Co-Manipulation in Leader-Follower


We aim to collaboratively handle heavy objects using a multi-robot system. By implementing a Leader-Follower control mechanism, the Follower, which is unaware of the target path, actively assists in the collaboration. With the appropriate assistance from the Follower, the Leader robot can perform the intended task with less force.