Experimental evaluation of joint position estimation using inertial sensors

In this paper we study the problem of estimating the position of a joint that is connecting two rigid links in a biomechanical model. By equipping the two links with inertial sensors, which measure linear acceleration and angular velocity, it is possible to estimate the joint position. Estimation methods for this problem have been proposed before, but experimental evaluation and comparison between methods is lacking. The main contribution of this paper is an experimental evaluation of three different methods, a least-squares method and two iterative optimization methods that minimize a sum of squared errors and a sum of absolute errors. The iterative methods turns out to be superior to the least-squares method. Minimizing the sum of squared errors is faster but not as robust against outliers as minimizing the sum of absolute errors.