Section: Overall Objectives

Introduction

Main challenge: The overall objective of the Team-Project e-Motion is to address some fundamental and open issues located at the heart of the emerging research field called “Human Centered Robotics’’. More precisely, our goal is to develop Perception, Decision, and Control algorithmic models whose characteristics fit well with the constraints of human environments; then, these models have to be embedded into “artificial systems” having the capability to evolve safely in human environments while having various types of interactions with human beings. Such systems have to exhibit sufficiently efficient and robust behaviors for being able to operate in open and dynamic environments, i.e. in partially known environments, where time, dynamics and interactions play a major role. Recent technological progress on embedded computational power, on sensor technologies, and on miniaturized mechatronic systems, make the required technological breakthroughs potentially possible (including from the scalability point of view).

Approach and research themes: Our approach for addressing the previous challenge is to combine the respective advantages of the Computational Geometry and of the Theory of Probabilities, while working in cooperation with neurophysiologists for the purpose of taking into account Human perception and navigation models. Two main research themes are addressed under both the algorithmic and human point of views; these research themes are respectively related to the problems of understanding dynamic scenes in human environments and of navigating interactively and safely in such environments.

• Perception & Situation awareness in Human environments. The main problem is to understand complex dynamic scenes involving human beings, by exploiting prior knowledge and a flow of perceptive data coming from various sensors. Our approach for solving this problem is to develop three complementary paradigms:

  • Bayesian Perception: How to take into account prior knowledge and uncertain sensory data in a dynamic context ?

  • Risk Assessment: How to evaluate this collision risk (i.e. potential future collisions) from an estimate of the current state of the dynamic scene, and from the prediction of the future behaviors of the scene participants ?

  • Behavior modeling & Learning: How to model and to learn behaviors from observations ?

• Navigation, Control, and Interaction in Human environments. The main problem is to take safe and socially acceptable goal-oriented navigation and control decisions, by using prior knowledge about the dynamic scenario and the related social rules, and by fusing noisy sensory data in order to estimate the state parameters. Our approach for addressing this problem is to develop two complementary concepts:

  • Human-Aware Navigation: How to navigate safely towards a given goal in a dynamic environment populated by human beings, while taking into account human-robot interactions and while respecting social rules and human comfort ?

  • State Estimation & Control: How to estimate the state parameters from noisy and sometime missing sensory data ? How to control a robot or a fleet of robots for executing a task in a near optimal way ?