Combining reinforcement learning and differential inverse kinematics for collision-free motion of multilink manipulators | Publicación