Seminar: Paolo Salaris, "From optimal synthesis to optimal visual servoing for autonomous vehicles", July 4 at 12:15
Speaker: Paolo Salaris, Università di Pisa
Title: From optimal synthesis to optimal visual servoing for autonomous vehicles
Time and location: Monday, July 4, 2011, at 12:15, Aula Magna
Abstract: Thanks to well-established advances in point-feature extraction and tracking algorithms, visual control is getting widespread in robotics. However, few practical problems still affect visual servoing approaches and depend on the particular available robotic set-up. For example, in case of limited Field-Of-View (FOV) cameras, the problem is of maintaining in sight the features necessary for the visual servoing during the robot manoeuvres. In this talk I present the problem of visual servo control for a unicycle-like vehicle equipped with a monocular fixed vision system. The system, subject to nonholonomic constraints imposed by the vehicle kinematics and to FOV constraints imposed by camera, must reach a desired position on the motion plane following the optimal (shortest) path. I present a general synthesis of shortest paths considering also the case of side-looking sensor systems where the forward direction is not necessarily included inside the sensor FOV. This solution is relevant also for underwater robotics application such surveying and navigation, where Autonomous Underwater Vehicles are equipped with side sonar scanners. The approach used to obtain the global partition of the motion plane induced by shortest paths is based on the exploitation of geometric symmetries and invariants in order to obtain a synthesis first for the points on the border of a compact subset of the motion space then for the interior of this subset and finally extended to the entire motion plane. Once the optimal synthesis is available, a crucial step towards the practical application of the shortest path synthesis is to translate the optimal trajectories into feedback control laws, as a function of the current state of the system only. In this talk, I present such feedback control laws and I prove stability properties for the proposed control scheme in a properly generalized analysis setting, which is that of stability on a manifold, by using the LaSalle's invariance principle. Finally, based on a visual control scheme, simulation and experimental results will be shown to prove the effectiveness of the proposed technique.