Vision-aided inertial navigation using virtual features

In this paper we consider an aerial vehicle equipped with a monocular camera and inertial sensors. Additionally, a laser pointer is mounted on the vehicle and it produces a laser spot. The laser spot is observed by the monocular camera and it is the unique point feature used in the proposed approach. We focus our attention to the case when the vehicle moves in proximity of a planar surface and in particular when the laser spot belongs to this surface. The paper provides two main contributions. The former is the analytical derivation of all the observable modes, i.e. all the physical quantities that can be determined by only using the inertial data and the camera observations of the laser spot during a short time-interval. Specifically, it is shown that the observable modes are: the distance of the vehicle from the planar surface; the component of the vehicle speed, which is orthogonal to the planar surface; the relative orientation of the vehicle with respect to the planar surface; the orientation of the planar surface with respect to the gravity. The second contribution is the introduction of a simple recursive method to perform the estimation of all the aforementioned observable modes. This method is based on a local decomposition of the original system, which separates the observable modes from the rest of the system. The method is validated by using synthetic data. Additionally, preliminary tests with real data are provided and more complete experiments are in progress. The presented approach can be integrated in the framework of autonomous take-off and landing, safe touch-down and low altitude manoeuvres even in dark or featureless environment.