Deterministic Learning-based Impedance Control for Human-Robot Interaction with Nonlinear Uncertain Dynamics

This is the authors' abstract. We don't add key points for this paper.

In this paper, we propose a Deterministic Learning-based impedance control scheme for human-robot interaction (HRI), which can deal with the uncertain nonlinear dynamics resulted by the unpredictable behavior in HRI process. A dynamicslearning-based impedance controller is proposed by embedding an adaptive radial basis function neural network (RBF NN) model to online learn the uncertain nonlinear dynamics of the robot from the closed-loop control during HRI process. With the proposed approach, the uncertain nonlinear dynamics of the robot can be accurated learned by guaranteeing the NN weights to converge to their optimal values, and the associated knowledge can be obtained and stored by a constant RBF NN model. A new experience-based controller is then developed with the constant NN model, which can recall the stored knowledge whenever the same/similar control situation recurs, thereby leadning to high control performance with less computation efforts.