Stability analysis and robust control of human intention-based physical human-robot interaction.


Access rights

No access – contact

Journal Title

Journal ISSN

Volume Title



Two main challenges that need to be addressed in physical human-robot interaction (pHRI) are efficient recognition of human intention and interaction safety. Human intention adaptation is usually realized by changing admittance parameters according to human interaction in physical human-robot interaction (pHRI). However, admittance parameters inferred from human intention may result in instability. This dissertation conducts a fundamental and systematic study on variable admittance control for assistive pHRI considering human intention adaptation, system passivity, and system stability. Human-intention-governed variable admittance control (VAC) is applied under a general human intention framework to shape the mechanical admittance to desired interaction. For the changing trends of these parameters (i.e., increasing or decreasing), those impacting system passivity are studied. A power envelope regulation (PER) concept is then proposed to impose constraints on variable admittance parameters inferred from human intention to maintain safe interaction and to preserve system passivity. It allows drastic changes in admittance controller dynamics, which usually result in instability, to be restrained. Our initial results suggest that the passivity condition is a necessary condition for system stability in pHRI. To yield stable and robust performance of VAC, a new sliding mode control (SMC) strategy is proposed. A universal sliding surface, regardless of the order of admittance equation, is defined first, and its validation in realizing desired variable admittance is theoretically proved. The reachability and robustness of the proposed controller are proved based on Lyapunov stability for VAC. Then, a trade-off between chattering removal and tracking performance is achieved by developing a new variable-boundary approach. Furthermore, acceleration feedback is applied to the proposed controller to improve robustness and tracking performance further. For human-leading case, the developed method in this dissertation yields an average 0.0035 m/s deviation on velocity tracking with 7% outliers (maximum deviation is 0.021 m/s) compared with 0.0664 m/s for conventional SMC method. The robustness of the proposed controller is verified by modifying the modeling error of the most distal link’s mass higher than 200%. The effectiveness of the proposed methods and the theoretical derivation are validated via numerical simulation and experiments on a manipulator in a 3-degree-of-freedom (DoF) planar configuration.