Application of the controlled active vision framework to robotic and transportation problems

摘要: Flexible operation of a robotic agent in an uncalibrated environment requires the ability to recover unknown or partially known parameters workspace through sensing. Of sensors available agent, visual provide information that is richer and more complete than other sensors. In this paper we present robust techniques for derivation depth from feature points on target's surface accurate high-speed tracking moving targets. We use these system operates with little no priori knowledge object-related environment. The designed under controlled active vision framework robustly determines such as velocity objects maps depths structure. Such determination intrinsic environmental essential performing higher level tasks inspection, exploration, grasping, collision-free motion planning. For both applications, Minnesota Robotic Visual Tracker (a single sensor mounted end-effector manipulator combined real-time system) automatically select surfaces, derive estimate parameter question, apply control vector based upon estimates guide manipulator. concludes applications transportation problems vehicle tracking. >