Teaching physical collaborative tasks: object-lifting case study with a humanoid

Abstract

Abstract-This paper presents the application of a statistical framework that allows to endow a humanoid robot with the ability to perform a collaborative manipulation task with a human operator. We investigate to what extent the dynamics of the motion and the haptic communication process that takes place during physical collaborative tasks can be encapsulated by the probabilistic model. This framework encodes the dataset in a Gaussian Mixture Model, which components represent the local correlations across the variables that characterize the task. A set of demonstrations is performed using a bilateral coupling teleoperation setup; then the statistical model is trained in a pure follower/leader role distribution mode between the human and robot alternatively. The task is reproduced using Gaussian Mixture Regression. We present the probabilistic model and the experimental results obtained on the humanoid platform HRP-2; preliminary results assess our theory on switching behavior modes in dyad collaborative tasks: when reproduced with users which were not instructed to behave in either a follower or a leader mode, the robot switched automatically between the learned leader and follower behaviors.

Motion learning and adaptive impedance for robot control during physical interaction with humans

“…5-(a),(b); note that the data exhibit a force-velocity correlation, this correlation is similar to the one observed in real-world data acquired with the HRP-2 robot (see [2]). Importantly, the force-velocity dependency defines a task model: how to generate a motion that is consistent with perceived force; i.e.…”

“…During task execution, the robot-follower adapts its desired stiffnessK d and inertiã Λ d , so as to ensure accurate reproduction of a learned task model. previous research where we proposed a learning algorithm to control the motion of the HRP-2 robot engaged in a similar task with a human [1], [2]. These experiments have revealed how important it is to be able to anticipate the forces (applied at the robot's end-effector), so as to generate an appropriate E. Gribovskaya and A. Billard are with Learning Algorithms and Systems Laboratory (LASA), EPFL, Switzerland; {elena.gribovskaya,aude.billard}@epfl.ch.…”

Prediction of Intention during Interaction with iCub with Probabilistic Movement Primitives

“…In Evrard et al (2009), they use GMMs and Gaussian Mixture Regression to learn, in addition to the position (joint information), force information. Using this method, a humanoid robot is able to collaborate in one dimension with its partner for a lifting task.…”

Robot learning from demonstration of force-based tasks with multiple solution trajectories

“…This implies that the learning framework's goal is not to learn merely a trajectory [14] or a task with predefined states as in assembly processes [19] that can be represented at a symbolic level. For endowing the robot with a suitable learning structure for this kind of tasks and avoiding to assume some aspects about the task to be learned, we propose to use a HMM to encode the teacher demonstrations using an ergodic topology, similar to the approach followed in [17].…”

“…Based on the former considerations, recent works have focused their efforts on exploiting this new data source in LbD. [14] proposed a learning framework based on Gaussian Mixture Model (GMM) and Gaussian Mixture Regression (GMR) for endowing a humanoid robot with the ability to perform a collaborative manipulation task with a human operator using a haptic device and working at trajectory level (just a simple vertical movement was learned). An extension of this research [15], combines LbD and adaptive control for teaching the task, which endows the robot with variable inertia and an adaptive algorithm to generate different reference kinematic profiles depending on the perceived force.…”