Extended Kalman filtering for flexible joint space robot control

Abstract

In this paper, an extended Kalman filter (EKF) strategy to estimate state variables from noisy measurements in flexible joint space manipulators is presented. First, an EKF that estimates the link and motor positions/velocities using only measurements from motor sensors is developed for space robots modeled with a classical linear joint dynamics model. Second, an extension for a novel nonlinear joint dynamics formulation is provided. The state estimates are coupled to a flexible joint adaptive controller in order to provide a complete closed-loop solution for real-time estimation and control. In numerical simulations, the EKF-adaptive controller combination demonstrates, for both dynamics representations, good performance when tracking a 12.6 × 12.6 m square trajectory.