Aude Billard

On Learning, Representing, and Generalizing a Task in a Humanoid Robot

We present a programming-by-demonstration framework for generically extracting the relevant features of a given task and for addressing the problem of generalizing the acquired knowledge to different contexts. We validate the architecture through a series of experiments, in which a human demonstrator teaches a humanoid robot simple manipulatory tasks. A probability-based estimation of the relevance is suggested by first projecting the motion data onto a generic latent space using principal component analysis. The resulting signals are encoded using a mixture of Gaussian/Bernoulli distributions (Gaussian mixture model/Bernoulli mixture model). This provides a measure of the spatio-temporal correlations across the different modalities collected from the robot, which can be used to determine a metric of the imitation performance. The trajectories are then generalized using Gaussian mixture regression. Finally, we analytically compute the trajectory which optimizes the imitation metric and use this to generalize the skill to different contexts

Learning and Reproduction of Gestures by Imitation

We presented and evaluated an approach based on HMM, GMR, and dynamical systems to allow robots to acquire new skills by imitation. Using HMM allowed us to get rid of the explicit time dependency that was considered in our previous work [12], by encapsulating precedence information within the statistical representation. In the context of separated learning and reproduction processes, this novel formulation was systematically evaluated with respect to our previous approach, LWR [20], LWPR [21], and DMPs [13]. We finally presented applications on different kinds of robots to highlight the flexibility of the proposed approach in three different learning by imitation scenarios.

Learning Stable Nonlinear Dynamical Systems With Gaussian Mixture Models

This paper presents a method to learn discrete robot motions from a set of demonstrations. We model a motion as a nonlinear autonomous (i.e., time-invariant) dynamical system (DS) and define sufficient conditions to ensure global asymptotic stability at the target. We propose a learning method, which is called Stable Estimator of Dynamical Systems (SEDS), to learn the parameters of the DS to ensure that all motions closely follow the demonstrations while ultimately reaching and stopping at the target. Time-invariance and global asymptotic stability at the target ensures that the system can respond immediately and appropriately to perturbations that are encountered during the motion. The method is evaluated through a set of robot experiments and on a library of human handwriting motions.

Learning human arm movements by imitation: Evaluation of a biologically inspired connectionist architecture

Reference LSA3-CONF-2000-001 URL: http://humanoids.usc.edu Record created on 2005-11-16, modified on 2017-05-10

A survey of Tactile Human-Robot Interactions

Robots come into physical contact with humans in both experimental and operational settings. Many potential factors motivate the detection of human contact, ranging from safe robot operation around humans, to robot behaviors that depend on human guidance. This article presents a review of current research within the field of Tactile Human-Robot Interactions (Tactile HRI), where physical contact from a human is detected by a robot during the execution or development of robot behaviors. Approaches are presented from two viewpoints: the types of physical interactions that occur between the human and robot, and the types of sensors used to detect these interactions. We contribute a structure for the categorization of Tactile HRI research within each viewpoint. Tactile sensing techniques are grouped into three categories, according to what covers the sensors: (i) a hard shell, (ii) a flexible substrate or (iii) no covering. Three categories of physical HRI likewise are identified, consisting of contact that (i) interferes with robot behavior execution, (ii) contributes to behavior execution and (iii) contributes to behavior development. We populate each category with the current literature, and furthermore identify the state-of-the-art within categories and promising areas for future research.

Discovering Optimal Imitation Strategies

This paper develops a general policy for learning the relevant features of an imitation task. We restrict our study to imitation of manipulative tasks or gestures. The imitation process is modeled as a hierarchical optimization system, which minimizes the discrepancy between two multi-dimensional datasets. To classify across manipulation strategies, we apply a probabilistic analysis to data in Cartesian and joint spaces. We determine a general metric that optimizes the policy of task reproduction, following strategy determination. The model successfully discovers strategies in six different imitative tasks and controls task reproduction by a full body humanoid robot.

Discriminative and Adaptive Imitation in Uni-Manual and Bi-Manual Tasks

This paper addresses the problems of what to imitate and how to imitate in simple uni- and bi-manual manipulatory tasks. To solve the what to imitate issue, we use a probabilistic method, based on Hidden Markov Models, for extracting the relative importance of reproducing either the gesture or the specific hand path in a given task. This allows us to determine a metric of imitation performance. To solve the how to imitate issue, we compute the trajectory that optimizes the metric, given a set of robots body constraints. We validate the methods in a series of experiments, where a human demonstrator teaches through kinesthetic a humanoid robot how to manipulate simple objects.

Dynamical System Modulation for Robot Learning via Kinesthetic Demonstrations

We present a system for robust robot skill acquisition from kinesthetic demonstrations. This system allows a robot to learn a simple goal-directed gesture and correctly reproduce it despite changes in the initial conditions and perturbations in the environment. It combines a dynamical system control approach with tools of statistical learning theory and provides a solution to the inverse kinematics problem when dealing with a redundant manipulator. The system is validated on two experiments involving a humanoid robot: putting an object into a box and reaching for and grasping an object.

Building Robota, a Mini-Humanoid Robot for the Rehabilitation of Children with Autism

The Robota project constructs a series of multiple-degrees-of-freedom, doll-shaped humanoid robots, whose physical features resemble those of a human baby. The Robota robots have been applied as assistive technologies in behavioral studies with low-functioning children with autism. These studies investigate the potential of using an imitator robot to assess childrens imitation ability and to teach children simple coordinated behaviors. In this article, the authors review the recent technological developments that have made the Robota robots suitable for use with children with autism. They critically appraise the main outcomes of two sets of behavioral studies conducted with Robota and discuss how these results inform future development of the Robota robots and robots in general for the rehabilitation of children with complex developmental disabilities.

Effects of repeated exposure to a humanoid robot on children with autism

This work is part of the Aurora project which investigates the possible use of robots in therapy and education of children with autism (Aurora, 2003), based on findings that people with autism enjoy interacting with computers, e.g. (Powell, 1996). In most of our trials we have been using mobile robots, e.g. (Dautenhahn and Werry, 2002). More recently we tested the use of a humanoid robotic doll. In (Dautenhahn and Billard, 2002) we reported on a first set of trials with 14 autistic subjects interacting with this doll. In this chapter we discuss lessons learnt from our previous study, and introduce a new approach, heavily inspired by therapeutic issues. A longitudinal study with four children with autism is presented. The children were repeatedly exposed to the humanoid robot over a period of several months. Our aim was to encourage imitation and social interaction skills. Different behavioural criteria (including Eye Gaze, Touch, and Imitation) were evaluated based on the video data of the interactions. The chapter exemplifies the results that clearly demonstrate the crucial need for long-term studies in order to reveal the full potential of robots in therapy and education of children with autism.