Learning bimanual end-effector poses from demonstrations using task-parameterized dynamical systems

Abstract

Very often, when addressing the problem of human-robot skill transfer in task space, only the Cartesian position of the end-effector is encoded by the learning algorithms, instead of the full pose. However, orientation is just as important as position, if not more, when it comes to successfully performing a manipulation task. In this paper, we present a framework that allows robots to learn the full poses of their end-effectors in a task-parameterized manner. Our approach permits the encoding of complex skills, such as those found in bimanual manipulation scenarios, where the generalized coordination patterns between end-effectors (i.e. position and orientation patterns) need to be considered. The proposed framework combines a dynamical systems formulation of the demonstrated trajectories, both in R 3 and SO(3), and taskparameterized probabilistic models that build local task representations in both spaces, based on which it is possible to extract the relevant features of the demonstrated skill. We validate our approach with an experiment in which two 7-DoF WAM robots learn to perform a bimanual sweeping task.

Learning Physical Collaborative Robot Behaviors From Human Demonstrations

“…For simplicity of the experiment, the orientation of the robot end-effector was kept constant. However, the approach also supports tasks with variable end-effector orientations [49].…”

Learning Task Priorities from Demonstrations

“…In previous work, task parameters have been used to represent poses of objects in a robot workspace, mapping local models of demonstrations (from the perspective of P different objects) onto a global coordinate system, typically the robot base frame. In this case, A (j) is a rotation matrix [8], [10], [37], [38] or a quaternion matrix [9], representing an object orientation, and b (j) is a translation vector, representing the origin of an object coordinate system with respect to the base frame of the robot. It was common in previous work to refer to task parameters as candidate frames or candidate coordinate systems.…”

“…For this reason, most research on bimanual skill learning exploits operational space formulations (e.g. [3], [4], [5], [6], [7], [8], [9]), that focus on task space constraints, e.g. demonstrated coordination between end-effectors and object-related movements, that need to be reproduced precisely in order to successfully complete a task.…”

Learning from demonstration for semi-autonomous teleoperation

“…We use unit quaternions to represent orientation data, as in Silvério et al (2015). We chose this representation as it is singularity-free and uses only 4 parameters.…”

“…This way we can directly use this representation of orientations under the TP-HSMM formulation by setting ξ n =ˆ n , b n, j = 0 and A n, j = E n, j for the orientation part of the task parameters, see Silvério et al (2015) for details. The projection of Eq.…”