Estimating the orientation of a rigid body moving in space using inertial sensors

This paper presents a novel method of estimating the orientation of a rigid body moving in space from inertial sensors, by discerning the gravitational and inertial components of the accelerations. In this method, both a rigid-body kinematics model and a stochastic model of the human-hand motion are formulated and combined in a nonlinear state-space system. The state equation represents the rigid body kinematics and stochastic model, and the output equation represents the inertial sensor measurements. It is necessary to mention that, since the output equation is a nonlinear function of the state, the extended Kalman filter (EKF) is applied. The absolute value of the error from the proposed method is shown to be less than 5 deg in simulation and in experiments. It is apparently stable, unlike the time-integration of gyroscope measurements, which is subjected to drift, and remains accurate under large accelerations, unlike the tilt-sensor method.