• Vehicle Control
• Visual Odometry
• Structure from Motion
• Simultaneous Localization and Mapping
• Nonlinear Observer
• Nonlinear Observability
• Nonlinear Control
• Motion Reconstruction
• Visual Servoing

In this thesis we investigate how monocular image measurements can be used as the single source of information for a vehicle to sense and navigate through its surroundings.

First we investigate what is the subset of vehicle location, environment mapping and vehicle motion that can be retrieved from images only. In particular, results apply to the case where no model of the vehicle, nor odometry or acceleration measurements are available. Then, we investigate the use of the information that can be extracted from images on visual servoing tasks and we define a servoing approach, named {\em Appearance Servoing}, that explicitly imposes the existing control constraints in the navigation of an Appearance Map. Finally, we present an experimental study case of the use of appearance where a sequence of images is used to construct a simple topological map of an office environment and then navigate a robot within it.