Artificial neural network based compliant control for robot arms

Abstract

The aim of this paper is to present an artificial neural network (ANN) based adaptive nonlinear control approach of a robot arm, with highlight on its capability as a compliant control scheme. The approach is based on a computed torque law and consists of two main components: A feedforward controller (approximated by the ANN) and a proportional-derivative (PD) feedback loop. Here, the feedforward controller is used to approximate the nonlinear system dynamics and can also adapt to the long-term dynamics of the arm while the PD feedback loop can be tuned to obtain proper compliant behaviour to deal with instantaneous disturbances (e.g., collisions). The employed controller structure makes it possible to decouple these two components for individual parameter adjustments. The performance of the control approach is evaluated and demonstrated in physical simulation which shows promising results.