We aim to develop an algorithm for robots to manipulate novel objects as tools for
completing different task goals. An efficient and informative representation would facilitate …

We discuss vision as a sensory modality for systems that interact flexibly with uncontrolled
environments. Instead of trying to build a generic vision system that produces task …

In this paper we present a general, flexible framework for learning map**s from images to
actions by interacting with the environment. The basic idea is to introduce a feature-based …

Abstract Approximate Policy Iteration (API) is a reinforcement learning paradigm that is able
to solve high-dimensional, continuous control problems. We propose to exploit API for the …

In der Einleitung wird zunächst das in dieser Arbeit behandelte visuomotorische
Lernproblem motiviert. Von der biologischen Lösung beim Menschen werden drei …

In the research to realize high standard task-oriented assistant robots, a general and
strategic way of development is essential. Otherwise high functionality and potential for …

We describe two quite different methods for associating action parameters to visual
percepts. Our RLVC algorithm performs reinforcement learning directly on the visual input …

Dee, HM [Hannah M.], Velastin, SA [Sergio A.], How close are we to solving the problem of
automated visual surveillance?: A review of real-world surveillance, scientific progress and …

I would like to thank Monique Thonnat the head of ORION team for her support, competence
and help during all these years to accomplish this HDR. She gave me the opportunity to …