Toward Orientation Learning and Adaptation in Cartesian Space

As a promising branch of robotics, imitation learning emerges as an important way to transfer human skills robots, where demonstrations represented in Cartesian or joint spaces are utilized estimate task/skill models that can be subsequently generalized new situations. While positions suffices for many applications, the end-effector orientation is required others. Despite recent advances orientations from demonstrations, several crucial issues have not been adequately addressed yet. For instance, how demonstrated adapted pass through arbitrary desired points comprise and angular velocities? In this article, we propose approach capable multiple trajectories adapting learned situations (e.g., via-points end-points), both velocity considered. Specifically, introduce kernelized treatment alleviate explicit basis functions when orientations, which allows associated with high-dimensional inputs. addition, extend our quaternions acceleration jerk constraints, generating smoother profiles robots. Several examples including experiments real 7-DoF robot arms provided verify effectiveness method.