SPIDAR calibration using Support Vector Regression

Abstract

This paper aims to present all the study done on the SPIDAR, which is a tracking and haptic device, in order to improve its accuracy on the given position. Firstly we proposed a new semi-automatic initialization technique for this device using an optical tracking system. Then, we propose to use Support Vector Regression (SVR) to calibrate the SPIDAR in order to reduce location errors. We obtained very good results with this calibration, since we reduced the mean error by more than 50%.