I. Introduction
In recent years, there has been growing interest in methods for motion estimation using visual and inertial measurements, a task often termed termed vision-aided inertial navigation (see, e.g. [1]–[7] and references therein). Many reasons exist for this. First, both cameras and MEMS inertial measurement units (IMUs) are compact, inexpensive, and have low power requirements. Second, these sensors can operate in virtually any environment, and allow for full-3D pose estimation, thus providing a very versatile solution for navigation. Third, in recent years we have seen a proliferation of devices (e.g., mobile phones) that contain both cameras and inertial sensors as part of their standard sensor payload. These devices are gradually becoming ubiquitous, and necessitate new techniques that will allow high-precision navigation in GPS-denied environments.