Active stability observer using artificial neural network for intuitive physical human–robot interaction

Abstract

Physical human–robot interaction may present an obstacle to transparency and operations’ intuitiveness. This barrier could occur due to the vibrations caused by a stiff environment interacting with the robotic mechanisms. In this regard, this article aims to deal with the aforementioned issues while using an observer and an adaptive gain controller. The adaptation of the gain loop should be performed in all circumstances in order to maintain operators’ safety and operations’ intuitiveness. Hence, two approaches for detecting and then reducing vibrations will be introduced in this study as follows: (1) a statistical analysis of a sensor signal (force and velocity) and (2) a multilayer perceptron artificial neural network capable of compensating the first approach for ensuring vibrations identification in real time. Simulations and experimental results are then conducted and compared in order to evaluate the validity of the suggested approaches in minimizing vibrations.