This work addresses the problem of visual control of mobile robots. We consider a vision system consisting of a monocular camera onboard the vehicle. We also consider the specific problem of pure visual control, therefore the approaches investigated are only based on visual information, and no odometry or additional sensors are needed. In favour of versatility, methods that do not require any special modelling of the environment and those that do not consider the use of fiducial markers are of significant importance. Then, we add to the framework the requirement of dealing with unknown and unstructured environments.

Following this idea only a snapshot image of the world taken at the goal location is given to the control system in order to perform the autonomous navigation task. We tackle the problem of visual control taking advantage of the multiple view geometry. Thus, the image information extracted is filtered through the constraints imposed by the geometric model across the views resulting in a better performance and robustness. Besides, we focus on the multiple view geometry itself (The epipolar geometry, homography or trifocal tensor), and we have proposed to solve the visual control task directly in terms of the geometric model.