Rigid body state estimation using dynamic vision and inertial sensors

In this paper, estimation of rigid body orientation is studied. The measurements are angular velocities from a rate gyro and the perspective projections of lines, obtained from a computer vision system. The problem is formulated as an implicit output problem for a system evolving on the special orthogonal group and it is shown that it is possible to decouple the rotation estimation problem from the position estimation. Observability is studied for different line configurations and an observer for the rotation is proposed. It is shown that under some natural conditions on the translation, the observer is locally exponentially stable. The observer does not use any local parameterization of the special orthogonal group but instead evolves on the group of rotation matrices.