Observer-Based Sensorless Admittance Control from Demonstrations with Dynamical Systems for Robot-Environment Interaction

Compliant motion adaptation is essential in the task of robot interacting with the environment. This paper presents a motion planning framework that generates and tracks a trajectory to achieve the compliant behavior during interaction. First, motions from multiple demonstrations are modelled by a dynamical system (DS) with Gaussian Mixture Models (GMM), and parameters of DS model are learning online. In case of the contacting force from the environment, admittance control is employed to modify the trajectory to generate a compliant robot motion. The contacting force is estimated by the sensorless approach. Furthermore, in our framework, an adaptive controller is designed to ensure the tracking of the modelled DS motion, and Radial Basis Function neural networks (RBFNNs) are adopted to compensate nonlinear terms of the robot. The proposed framework is not only analyzed in Lyapunov theory, and also verified through simulations study.