Towards Robust Physical Human Robot Interaction by an Adaptive Admittance Controller

Abstract

The regular admittance controller cannot be easily transferred to the physical human-robot interaction scenario because of the dynamic stiffness of the human arm. The dynamic interaction of humans can cause high frequency and unsafe oscillation of the robot arm. Based on the adaptive control scheme, this paper presents an online sensory-based analytical approach to recognize and quantify the “stability index” named as a robust haptic observer. The observer performs the Fast Fourier Transform on the interaction force signal within a sliding window and quickly detects system oscillation through a simple mathematical transformation. Compared with the existing methods, it can calculate a normalized system stability index more accurately and faster. This quantified index is employed in a linearized adaptive law to tune the parameters of the admittance controller. Experimental validation of the proposed strategy is performed and compared with state-of-the-art work in a task of human-guided drawing. The results show that our proposed approach can effectively detect oscillation, and the drawing time is shortened by 15% with the same tracking accuracy. In addition, the energy consumption is 44.4% less on average.