New perspectives in robotics opened by high-speed vision and rolling shutter – VIRAGO

This project takes place in the framework of vision-based modeling, identification and control of parallel robots. Those points are now fairly well studied, both from a pure robotic point of view and from the one of geometric vision applied to parallel robots, in a kinematic approach. However, the dynamics associated to this problem remain largely unadressed in the international litterature. Following results obtained in 2005, showing that vision is unavoidable for the kinematic control of parallel robots (i.e. at lower speed), we want to bring the proof that it is also a necessity for dynamic control, that is at higher speed where inertias or Coriolis forces can not be neglected any more. To do so, we started last year to develop software tools for high-speed vision. It turned out that the camera we bought had a conceptual defect : it was grabbing the lines one after the other one with a rolling shutter, which can cause image distorsion due to a fast motion of the observed object. We turn this defect into a new ability, since on the basis of a novel perspective model we proposed, we can now not only estimate in a single image the pose but also the speed of a known object. This discovery, awarded at the higly selective European Conference on Computer Vision 2006 for its novelty, requires to be completed by an immediate and deepened study, if we do not want to loose our leadership on this field at the highest international level. This implies the setting up of a dedicated task force, which we propose to gather here. The scientific objectives opened by this discovery are therefore to take advantage of this new perspective model to : define a differential geometry never studied before ; propose a formal framework to high-speed-vision-based dynamic control of parallel robots ; design new kinds of high-speed vision sensors. To do so, the competences of several domains (robotics, computer vision, sensors) will have to be gathered, transversally with respect to the research group organisation. The strategic objective of this project is thus to create a team with sufficient forces, thanks to its multi-disciplinary character, to deal with the long term objectives above. Due to the width of the latter, this team should not reach all of them during this project and should hence keep on working on the field in the future. The scientific goals during the project duration recall the long term objectives: 1.geometry: application to high-speed metrology of parallel robots; 2.control : innovative schemes for high-speed vision-based control ; 3.sensors : design of an intelligent rolling shutter high-speed camera. The results will have to be demonstrated through the vision-based dynamic control of several prototypes, among which the Isoglide4-T3R1 robot jointly developped by LASMEA and LaMI.