This paper proposes a novel planning algorithm for image based visual servoing (IBVS). IBVS is conceptually simple but prone to fail in case of large displacements. Hence, the objective is to provide paths for reference points in the image plane, such that the camera is steered from initial position to desired position. The novel scheme guarantees that the camera is always oriented such that the object of interest is in the field of view and that any reference point set corresponds indeed to a feasible camera pose. No camera calibration is necessary. As a novelty, this work is planning a camera motion, hence yielding compact paths in robot work space. The proposed planner is compared to others by experiments.

We consider the problem of planning a path for image-based eye-in-hand visual servoing applications, when using an uncalibrated camera. Recent developments in this field lead to analytical interpolation between two views of an unknown object, in the projective space. Literature also provides extensions of these approaches that account for the target visibility constraint. Unfortunately, these extensions require an (even weakly) calibrated camera, and involve iterative algorithms that are not proven to converge. In this paper, we propose a modification of an existing projective interpolation algorithm that guarantees the visibility of the observed target, while avoiding the use of any knowledge on the camera calibration parameters, nor on the observed object

Trajectory planning for eye-in-hand visual servoing is usually performed in the Euclidean work space of the robot or in a two dimensional image space. However, planning in Euclidean space may lead to very inappropriate trajectories in image space and vice versa. These difficulties are due to the perspective transformation of the camera, loss of one dimension due to projection onto the image plane and the fact that only a rough approximation of camera parameters is practically available. Hence, this paper proposes a planning scheme for image trajectories such that straight line behaviour is ensured both in image space and world space, ie. a single but arbitrarily chosen point in the image plane performs straight line behaviour as well as the camera optical center in work space. This way, trajectories become very compact and most of the above mentioned problems are avoided in a natural way. The algorithm requires as a priori information nothing more than matched image points (in pixels) from a current and desired image and a depth set for at least one position.

This work deals with trajectory planning for image based eye-in-hand visual servoing applications. Trajectories for extracted feature points of a target object are computed directly in the projective image space such that they are compatible with rigid body displacements. To this end, one separately parameterizes for: (a) the collineation with respect to the plane at infinity (which can be viewed as a projective representation of the rotation); and (b) the projective representation of the translation. Parameterizing the collineation by the use of decomposition into its canonical form provides an easy way to plan for a geodesic displacement. Furthermore, different strategies are proposed to interpolate for the translation path. As a result, the most advanced strategy guarantees, by the use of a simple analytic criterion, that the points remain visible during the whole path. The method requires in general nothing more than the initial and final images and does not depend on intrinsic camera parameters.

This work presents a novel visual servoing approach, aimed at controlling the so-called extended-2D (E2D) coordinates of the points constituting a tracked target. This approach does not require any pose reconstruction. Rather, the only information required to build the E2D coordinates are the estimated depth distribution of the target points, and the estimated camera model. Several implementations of the controller are considered, which are inspired from conventional image based visual servoing from points. In spite of their simplicity, the proposed control laws exhibit remarkable stability robustness properties. A key issue is that only three configurations of undesired equilibrium exist, and they are all proven to be unstable even in the uncalibrated case. In other words, contrarily to existing 2D methods, there are no local minima. The paper details the control design and analysis and provides simulation results, emphasizing the remarkable robustness with respect to camera calibration errors.