Coupled least-squares forward kinematics and extended Kalman filtering for the pose estimation of a cable-driven parallel robot

Abstract

This paper presents a novel pose estimation framework for a cable driven parallel robot (CDPR) built upon an extended Kalman filter (EKF) coupled with a nonlinear least-squares-based forward kinematics (FK) algorithm. This estimation approach makes use of an end-effector-mounted accelerometer and rate gyroscope, as well as measurements of the CDPR's cable lengths. Coupling the EKF with the dynamically-updating error covariance on the pose estimate from FK, results in a more accurate pose estimate than FK alone. Simulation results are presented that confirm this improved accuracy and filter consistency. Multiple experimental results are also included, which demonstrate that EKF is capable of providing much lower covariance on the estimation error (i.e., much greater confidence in the pose estimate) compared to FK alone.