Kinesthetic Learning Based on Fast Marching Square Method for Manipulation

The advancement of robotics in recent years has driven the growth robotic applications for more complex tasks requiring manipulation capabilities. Recent works have focused on adapting learning methods to which are stochastic and may not converge. In this paper, a kinesthetic method based fast marching square is presented. This poses great advantages such as ensuring convergence from experience human demonstrator. For purpose, demonstrator teaches paths by physically guiding one UR3 arms mobile manipulator. After first phase, Learning used make robot learn experience. As novelty, an auto-learning functionality presented, provides algorithm with exploration capacity. base only using information provided taught trajectories, but also expanding its ability order explore unknown states environment. effectiveness proposed been evaluated through simulations 2D 3D environments real process analyzed other approaches LASA dataset it tested scenarios different obstacles, proving effectiveness.