A Quaternion-Based Unscented Kalman Filter for Robust Optical/Inertial Motion Tracking in Computer-Assisted Surgery

Abstract

This paper presents a sensor fusion algorithm based on an unscented Kalman filter (UKF) designed for robust estimation of position and orientation of a freely moving target in surgical applications. The UKF is not subject to the nontrivial disadvantages of the more popular extended Kalman filter that can affect the accuracy or even lead to divergence of the algorithm. Orientation is represented by quaternion, which avoids singularities and is computationally more effective. The fusion algorithm has been designed to incorporate an optical tracking system and an inertial sensor unit containing triaxial angular rate sensors and accelerometers. The proposed tracking system does not suffer from environmental distortion, is robust to brief line-of-sight losses and has a high sampling rate. Experimental results validate the filter design, and show the feasibility of using optical/inertial sensor fusion for robust motion tracking meeting the requirements of surgical computer-assisted procedures.