A time-domain vibration observer and controller for physical human-robot interaction

•We propose a non-linear algorithm for measuring and eliminating vibrations coming from direct human-robot physical interaction.•We detect a signal’s minima and maxima with robustness in regards to noise.•We define a vibration index based on these extrema.•Based on this vibration index, we propose adaptive control strategies in order to reduce or eliminate the vibrations.•We implement the algorithm on two different prototypes in order to demonstrate its performance. [Display omitted] This paper presents a time-domain vibration observer and controller for physical Human-Robot Interaction (pHRI). The proposed observer/controller aims at reducing or eliminating vibrations that may occur in stiff interactions. The vibration observer algorithm first detects minima and maxima of a given signal with robustness in regards to noise. Based on these extrema, a vibration index is computed and then used by an adaptive controller to adjust the control gains in order to reduce vibrations. The controller is activated only when the amplitude of the vibrations exceeds a given threshold and thus it does not influence the performance in normal operation. Also, the observer does not require a model and can analyze a wide time frame with only a few computations. Finally, the algorithm is implemented on two different prototypes that use an admittance controller.