Les robots de service ou d'assistance doivent évoluer dans un environnent humain en constant changement, souvent imprévisible. Ils doivent donc être capables de s'adapter à ces changements, idéalement de manière autonome, afin de ne pas dépendre de la présence constante d'une supervision. Une telle adaptation en environnements non structurés nécessite notamment une détection et un apprentissage continu des nouveaux objets présents, que l'on peut imaginer inspirés des enfants, basés sur l'interaction avec leur parents et la manipulation motivée par la curiosité. Notre travail vise donc à concevoir une approche développementale permettant à un robot humanoïde de percevoir son environnement. Nous nous inspirons à la fois de la perception humaine en termes de fonctionnalités et du développements cognitifs observé chez les infants. Nous proposons une approche qui permet à un robot humanoïde d'ex- plorer son environnement de manière progressive, comme un enfant, grâce à des interactions physiques et sociales. Suivant les principes de la robotique développementale, nous nous concentrons sur l'apprentissage progressif, continu et autonome qui ne nécessite pas de connaissances a priori des objets. Notre système de perception débute par la segmentation de l'espace visuel en proto-objets, qui serviront d'unités d'attention. Chaque proto-objet est représenté par des caractéristiques bas-niveaux (la couleur et la texture) et sont eux-mêmes intégrés au sein de caractéristiques de plus haut niveau pour ensuite former un modèle multi-vues. Cet apprentissage s'effectue de manière incrémentale et chaque proto-objet est associé à une ou plusieurs entités physiques distinctes. Les entités physiques sont ensuite classés en trois catégories : parties du robot, parties des humains et objets. La caractérisation est basée sur l'analyse de mouvements des entités physiques provenant de la vision ainsi que sur l'information mutuelle entre la vison et proprioception. Une fois que le robot est capable de catégoriser les entités, il se concentre sur l'interaction active avec les objets permettant ainsi d'acquérir de nouvelles informations sur leur apparence qui sont intégrés dans leurs modèles de représentation. Ainsi, l'interaction améliore les connaissances sur les objets et augmente la quantité d'information dans leurs modèles. Notre système de perception actif est évalué avec le robot humanoïde iCub en utilisant une base expérimentale de 20 objets. Le robot apprend par interaction avec un partenaire humain ainsi que par ses propres actions sur les objets. Notre système est capable de créer de manière non supervisée des modèles cohérents des différentes entités et d'améliorer les modèles des objets par apprentissage interactif et au final de reconnaître des objets avec 88.5% de réussite.

This book addresses the central role played by development in cognition. The focus is on applying our knowledge of development in natural cognitive systems, specifically human infants, to the problem of creating artificial cognitive systems in the guise of humanoid robots. The approach is founded on the three-fold premise that (a) cognition is the process by which an autonomous self-governing agent acts effectively in the world in which it is embedded, (b) the dual purpose of cognition is to increase the agent's repertoire of effective actions and its power to anticipate the need for future actions and their outcomes, and (c) development plays an essential role in the realization of these cognitive capabilities. Our goal in this book is to identify the key design principles for cognitive development. We do this by bringing together insights from four areas: enactive cognitive science, developmental psychology, neurophysiology, and computational modelling. This results in roadmap comprising a set of forty-three guidelines for the design of a cognitive architecture and its deployment in a humanoid robot. The book includes a case study based on the iCub, an open-systems humanoid robot which has been designed specifically as a common platform for research on embodied cognitive systems .

A robot is a true blank slate, awash in sensory information inextricable from its own actions. As such, it can be a powerful tool for investigating the space of problems that an infant, or whatever innate machinery was granted to the infant by evolution, must solve. The key observation is that the environment in which an infant develops is the same as that in which a robot might exist. A robot may have a different view on that environment, through different types or qualities of sensors. A robot may have different capabilities for interacting with the environment, for example by having wheels instead of legs. But, to act autonomously and coherently in the world, like an infant learns to do, a robot must somehow make sense of its sensory information. The goal of this thesis, broadly, is to enable the pneumatic humanoid robot, Diego, to actively perceive the world. Specifically, three problems are addressed. How can a robot : (1) identify interesting information in its sensory input, (2) direct its sensors to best acquire meaningful information, (3) learn to generalize its experience to interact with novel objects? To help answer these questions, this thesis presents : (1) a model of salience that is suitable for active cameras, (2) a model of eye movements based on optimal control, and (3) a framework and robotic implementation for visual perception of the inertial properties of objects.

This book provides an overview of model-based environmental visual perception for humanoid robots. The visual perception of a humanoid robot creates a bidirectional bridge connecting sensor signals with internal representations of environmental objects. The objective of such perception systems is to answer two fundamental questions: What & where is it? To answer these questions using a sensor-to-representation bridge, coordinated processes are conducted to extract and exploit cues matching robot’s mental representations to physical entities. These include sensor & actuator modeling, calibration, filtering, and feature extraction for state estimation. This book discusses the following topics in depth: • Active Sensing: Robust probabilistic methods for optimal, high dynamic range image acquisition are suitable for use with inexpensive cameras. This enables ideal sensing in arbitrary environmental conditions encountered in human-centric spaces. The book quantitatively shows the importance of equipping robots with dependable visual sensing. • Feature Extraction & Recognition: Parameter-free, edge extraction methods based on structural graphs enable the representation of geometric primitives effectively and efficiently. This is done by eccentricity segmentation providing excellent recognition even on noisy & low-resolution images. Stereoscopic vision, Euclidean metric and graph-shape descriptors are shown to be powerful mechanisms for difficult recognition tasks. • Global Self-Localization & Depth Uncertainty Learning: Simultaneous feature matching for global localization and 6D self-pose estimation are addressed by a novel geometric and probabilistic concept using intersection of Gaussian spheres. The path from intuition to the closed-form optimal solution determining the robot location is described, including a supervised learning method for uncertainty depth modeling based on extensive ground-truth training data from a motion capture system. The methods and experiments are presented in self-contained chapters with comparisons and the state of the art. The algorithms were implemented and empirically evaluated on two humanoid robots: ARMAR III-A & B. The excellent robustness, performance and derived results received an award at the IEEE conference on humanoid robots and the contributions have been utilized for numerous visual manipulation tasks with demonstration at distinguished venues such as ICRA, CeBIT, IAS, and Automatica.

This book tries to address the following questions: How should the uncertainty and incompleteness inherent to sensing the environment be represented and modelled in a way that will increase the autonomy of a robot? How should a robotic system perceive, infer, decide and act efficiently? These are two of the challenging questions robotics community and robotic researchers have been facing. The development of robotic domain by the 1980s spurred the convergence of automation to autonomy, and the field of robotics has consequently converged towards the field of artificial intelligence (AI). Since the end of that decade, the general public’s imagination has been stimulated by high expectations on autonomy, where AI and robotics try to solve difficult cognitive problems through algorithms developed from either philosophical and anthropological conjectures or incomplete notions of cognitive reasoning. Many of these developments do not unveil even a few of the processes through which biological organisms solve these same problems with little energy and computing resources. The tangible results of this research tendency were many robotic devices demonstrating good performance, but only under well-defined and constrained environments. The adaptability to different and more complex scenarios was very limited. In this book, the application of Bayesian models and approaches are described in order to develop artificial cognitive systems that carry out complex tasks in real world environments, spurring the design of autonomous, intelligent and adaptive artificial systems, inherently dealing with uncertainty and the “irreducible incompleteness of models”.