Visual control for memory-based navigation using the trifocal tensor

Abstract

In this paper, we present a control scheme for visual path-following of wheeled mobile robots based on a robust geometric constraint: the trifocal tensor (TT). The proposed control law only needs one element of the TT as feedback information, which is computed from the current and the target images along the sequence of the visual path. The scheme is valid for images captured by cameras having approximately a unique center of projection, e.g., conventional, central catadioptric and some fisheye cameras. The benefits of the proposed scheme are that explicit pose parameters decomposition is not required and the rotational velocity is smooth or eventually piece-wise constant avoiding discontinuities that generally appear when a new target image must be reached. Additionally, the translational velocity is adapted as required for the path. The validity and performance of the approach is shown through realistic simulations using synthetic images.