Force-impedance control of a 2-DOF planar robot using model predictive control based on successive linearisation of neural network model

Abstract

This paper presents a position-based force-impedance controller for a class of robot manipulators. A multilayer perceptron (MLP) neural network (NN) based on the structure of a nonlinear autoregressive model with exogenous input (NARX) is used to accurately model the dynamics of the robot manipulator. Using the developed neural network-based prediction model, a position-tracking adaptive linear model predictive controller (MPC) is synthesised. The neural network model is trimmed and linearised successively about the reference trajectory in order to improve the performance. Finally, a force-impedance control loop is added to the position-tracking MPC to regulate the dynamic interaction between the end effector of the manipulator and the environment. The feasibility and effectiveness of the developed control scheme is verified through extensive simulation using a two degree of freedom (2-DOF) planar robot manipulator in MATLAB.