Personalization and Adaptation in Physical Human-Robot Interaction

Gopinathan S. Personalization and Adaptation in Physical Human-Robot Interaction. Bielefeld: Universität Bielefeld; 2019.Recent advancements in physical human-robot interaction (pHRI) makes it possible for compliant robots to assist the human counterpart while closely working together. An ideal control mode designed for pHRI should be easy to handle, intuitive to use, ergonomic and adaptive to human habits and preferences. The major stumbling block in achieving this is that each user has varying physical capabilities and characteristics. This variance in the user behavior and other features is often high and rather unpredictable, which hinders the development of such systems. To tackle this problem, the idea of personalized adaptive stiffness control for pHRI is introduced in this thesis. Extensive user-studies are conducted in scope of this thesis and various control modes for pHRI are proposed and evaluated using appropriate user-studies. Both naive and expert users were considered in the user-studies and inferences from each study were used to improve the control mode to be better suited for pHRI. The thesis follows a meticulous research plan, an initial user-study confirms the im- portance of pHRI and kinesthetic guidance in industrial tasks. Subsequently, the user interactive force based adaptation is proposed and a second user-study is conducted where it is compared with standard control modes for pHRI. Importance of task specific param- eters and the need for combining the task and human factors emerged from the results of the second user-study. In the next phase manipulability based approaches which com- bine both task and human parameters are proposed and validated by conducting a third user-study. In the final phase a fourth user-study is conducted where the proposed con- trol modes are compared against more complex methods that have been proposed in the literature. The importance of human physical factors and needs for human centered systems for pHRI is validated in this thesis. The results show that including these human factors not only improve the performance but also improves the interaction quality and reduces the complexity of the pHRI