Đang chuẩn bị liên kết để tải về tài liệu:
Learning Fine Motion in Robotics: Experiments with the Hierarchical Extended Kohonen Map

Computational fluid dynamics, usually abbreviated as CFD, is a branch of fluid mechanics that uses numerical methods and algorithms to solve and analyze problems that involve fluid flows. Computers are used to perform the calculations required to simulate the interaction of liquids and gases with surfaces defined by boundary conditions. With high-speed supercomputers, better solutions can be achieved. Ongoing research yields software that improves the accuracy and speed of complex simulation scenarios such as transonic or turbulent flows. Initial validation of such software is performed using a wind tunnel with the final validation coming in full-scale testing, e.g. flight tests | Learning Fine Motion in Robotics Experiments with the Hierarchical Extended Kohonen Map Cristina Versinof Luca Maria Gambardellaf f IDSIA Corso Elvezia 36 6900 Lugano Switzerland cristina@idsia.ch http www.idsia.ch cristina Abstract We present a Hierarchical Extended Kohonen Map HEKM and a planning system which cooperate to solve the robot path finding problem. The HEKM learns to associate appropriate actions to perceptions under the supervision of the planner. First we argue for the utility of using the hierarchical version of the KM instead of the flat KM the HEKM provides a natural and economic representation of the robot s perceptual states. Second we measure the benefits of cooperative learning due to the interaction of neighboring neurons in the HEKM with cooperation learning is slowed down in the short run but the benefits appear later on resulting in a more satisfactory final performance. Third we highlight a beneficial side-effect obtained by transferring motion skill from the planner to the HEKM namely smoothness of motion. 1 Introduction The problem of path finding has attracted considerable attention by robotics research. This is the problem of moving a robot from a starting position to a goal position avoiding collisions against obstacles in the workspace. Moreover the robot path should be as short and smooth as possible. The path is optimal if it is the shortest from the starting position to the goal. Path finders are methods to automatically solve the path finding problem. Traditionally path finders are either model-based or sensor-based. While model-based systems address the path finding problem globally using a model of the workspace sensor-based systems consider it locally and rely on robot sensors to avoid obstacles. Both methods have limitations which are rather complementary. Model-based systems compute optimal free-paths and recover easily from dead-ends but require a complete description of the robot workspace and are computationally .