Adding a Little Intelligence to Factory Robots

Traditional industrial robots show up without any software besides a compiler or interpreter for a robot motion language, and no external sensors or safety systems.  Systems integrators and consultants are used to define how the robot will operate in a production line, sensors and interfaces to other machines are procured, and a specific new program is written for the application.  In this talk I will describe a new way of thinking about industrial robots, where they came with sensors, machine interfaces, a User Interface, and in intelligence software stack that together provide most of the control that is needed to get the robot to do useful work.  Factory workers then show the robot what it is to do, in task coordinates rather than robot coordinates, and the robot is able to carry out the task and notice anomalous situations.  The robots are not very intelligent by the standards of AI researchers, but they are robust and reliable, and just a little intelligence makes them so much more powerful than traditional industrial robots.

(Image Le Seuil)

L’imaginaire de l’intelligence. 

L’intelligence est le mythe fondateur de l’informatique. De nos jours, elle semble nourrir tous les espoirs et toutes les utopies. Intelligence collective, Bille intelligente, Intelligence des données, Intelligence économique… Tout semble intelligent. Un retour vers les textes fondateurs de l’informatique moderne et un clin d’œil vers les romans de la science-fiction seront l’occasion d’un regard historique sur l’intelligence, les algorithmes et l’autonomie émergente des êtres numériques.

Une thèse en programmation par contraintes

La Programmation Par Contraintes est une discipline à la croisée de l'Intelligence Artificielle et de la Recherche Opérationnelle. On la retrouve dans de nombreuses applications d'ordonnancement et de planification, pour son caractère à la fois générique et performant. Néanmoins, le passage à l'échelle de telles approches pose parfois problème.

Dans cet exposé je présenterai quelques travaux de ma thèse, intitulée Exploitation de structures de graphe en programmation par contraintes. Le but de la thèse était double : mieux traiter les problèmes de graphe en PPC et améliorer les solveurs de contraintes grâce à des raisonnements de graphe. Cet exposé étant principalement destiné aux jeunes chercheurs, je présenterai également quelques éléments méthodologiques et points marquants de la thèse.

Progress and Challenges in Planning

Planning in AI represents the model-based approach to control where  actions are selected to achieve goals given a predictive model of the actions and sensors. The main challenges in planning are computational, as all the models, whether accommodating  feedback and uncertainty or not, are intractable in the general case. In this talk, I'll review some the models considered in planning,  and some recent ideas that have turned out to be useful computationally. I'll focus mainly on three types of methods from our own recent work: 1) width-based search, where we show results on a number of video games from the ALE and GVG-AI environments, 2) translations for planning with partial observability  using classical planners, and 3) multiagent planning with nested beliefs. In all cases, we seek methods that can be both general and effective. This is joint work with students and collaborators, including Blai Bonet, Nir Lipovetzky, Miquel Ramirez, Filippos Kominis, and Tomas Geffner.

Preferences in Artificial Intelligence and Machine Learning

Reasoning with preferences has gained increasing attention in the field of artificial intelligence (AI) in the recent years. Amongst others, this interest has led to the emergence of preference learning as a new branch of machine learning. Roughly speaking, preference learning refers to the development and analysis of methods for learning preference models from explicit or implicit preference information, that is, models useful for explaining revealed and predicting future preferences. Starting with a brief introduction to preferences in AI in general and machine learning in particular, this talk will focus on methods and algorithms for ranking problems. These are (learning) problems in which preferences are represented in terms of a total order over an underlying set of choice alternatives, where both the order relation and the choice set can be contextualized. Based on different types of preference feedback, the goal of the learner is to induce a model that allows for predicting such order relations in an accurate manner.

Fair and Sustainable Resource Allocation in Self-Organising Multi-Agent Systems

Many open computing systems, for example grid and cloud computing, and ad hoc networks, such as sensor or vehicular networks, face a similar problem: how to collectivise resources, and distribute them fairly, in the absence of a centralized component. In this talk, we apply the methodology of sociologically-inspired computing, in which the study of human (social) models are formalised as the basis of engineering solutions to technical problems. In this case, we present formal models of Ostrom's design principles for self-governing institutions and Rescher's theory of distributive justice, for defining executable specifications of electronic institutions which support fair and sustainable resource allocation. This work has opened up a programme of research called computational justice: capturing some notions of `correctness' in the outcomes of algorithmic decision-making. However there are various different 'aspects' of justice, and some of these -- natural, distributive, retributive, procedural and interactional are discussed in the context of supporting self-governance for self-organising multi-agent systems.

Pérennisation de la connaissance via un Electronic Performance Support System (EPSS).

La société MBDA (Groupe EADS) fait partie du secteur de l’armement et emploie 10.000 personnes. Il y a 6 ans les questions de la pérennisation et de la capitalisation de la connaissance liée aux applications informatique ont été posées aux Responsable des Système d’Informatique et de la formation. Le retour d’expérience que je vous propose de partager correspond à la mise en place d’une base de connaissances pendant les 6 années qui ont suivi.

Je commencerai par la présentation du contexte et l'explicitation des questions posées après 10 ans de déploiement informatique par les salariés après une étude de satisfaction. J’aborderai la réflexion que nous avons eue sur la notion de connaissance, d’apprentissage et de compétence. Celle-ci nous a amené à la mise en place d’un EPSS (Electronic Performance Support System) que je définirai ainsi que l’organisation informatique, organisationnelle et de production que nous avons mise en place. Je ferai ensuite un bilan sur la bibliothèque de connaissance applicative métier créée pour finir par les perspectives envisageables dans les prochaines années.