João Silvério

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

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 ℝ3 and SO(3), and task-parameterized 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.

An Approach for Imitation Learning on Riemannian Manifolds

In imitation learning, multivariate Gaussians are widely used to encode robot behaviors. Such approaches do not provide the ability to properly represent end-effector orientation, as the distance metric in the space of orientations is not Euclidean. In this paper, we present an extension of common imitation learning techniques to Riemannian manifolds. This generalization enables the encoding of joint distributions that include the robot pose. We show that Gaussian conditioning, Gaussian product, and nonlinear regression can be achieved with this representation. The proposed approach is illustrated with examples on a two-dimensional sphere, with an example of regression between two robot end-effector poses, as well as an extension of task-parameterized Gaussian mixture model and Gaussian mixture regression to Riemannian manifolds.

Learning Controllers for Reactive and Proactive Behaviors in Human–Robot Collaboration

Designed to safely share the same workspace as humans and assist them in a variety of tasks, the new collaborative robots are targeting manufacturing and service applications that once were considered unattainable. The large diversity of tasks to carry out, the unstructured environments and the close interaction with humans call for collaborative robots to seamlessly adapt their behaviors so as to cooperate with the users successfully under different and possibly new situations (characterized, for example, by positions of objects/landmarks in the environment, or by the user pose). This paper investigates how controllers capable of reactive and proactive behaviors in collaborative tasks can be learned from demonstrations. The proposed approach exploits the temporal coherence and dynamic characteristics of the task observed during the training phase to build a probabilistic model that enables the robot to both react to the user actions and lead the task when needed. The method is an extension of the Hidden Semi-Markov Model where the duration probability distribution is adapted according to the interaction with the user. This Adaptive Duration Hidden Semi-Markov Model (ADHSMM) is used to retrieve a sequence of states governing a trajectory optimization that provides the reference and gain matrices to the robot controller. A proof-of-concept evaluation is first carried out in a pouring task. The proposed framework is then tested in a collaborative task using a 7 DOF backdrivable manipulator.