Using a Maximum Error Statistic to Evaluate Measurement Errors in 3D Position and Orientation Tracking Systems

Abstract

One approach to tracking anatomical and robot joint motion consists of tracking the XYZ locations of multiple point targets that are attached to each of the moving segments and then computing the three translations and three orientation angles between adjoining segments. The complexity of such systems requires that we introduce a new conservative maximum error statistic to be used for evaluating the accuracy of 3D motion tracking systems. This paper addresses the various phenomena that contribute to measurement error when computing six degrees of freedom associated with the relative motion between the adjacent segments. The characteristics of these errors, common to many 3D motion tracking systems, were first determined by experimentation using one such system (MnSCAN). These and additional artifacts were then modeled in order to quantitatively evaluate their effects using the maximum error statistic. Based on these computer experiments, several relationships were identified that predict how each of these phenomena influences the predicted measurement of relative motion between bodies. These suggest where design emphasis should be placed in order to minimize the error in tracking the six degrees of freedom. The methodology and the conclusions based on these results can be applied to designing most six degree of freedom position and motion measurement systems.