Joint axis and position estimation from inertial measurement data by exploiting kinematic constraints

Abstract

We consider 6d inertial measurement units (IMU) attached to rigid bodies, e.g. human limb segments or links of a robotic manipulator, which are connected by hinge joints and spheroidal joints. Novel methods for joint axis estimation and joint position estimation are presented that exploit the kinematic constraints induced by these two types of joints. The presented methods do not require any knowledge about the sensor units' positions or orientations and do not include integration, i.e. they are insensitive to measurement bias. By means of a three-links simulation model, the estimation algorithms are validated and convergence is analyzed. Finally, the algorithms are tested using experimental data from IMU-based human gait analysis.