Skip to Main content Skip to Navigation

Perception multimodale de l’homme pour l’interaction Homme-Robot

Christophe Mollaret 1
1 LAAS-RAP - Équipe Robotique, Action et Perception
LAAS - Laboratoire d'analyse et d'architecture des systèmes
Abstract : This thesis focuses on multimodal perception of man for Interaction Human-Robot (IHR). It was financed by the ANR restrained RIDDLE (2012-2015). This project focuses on the development of an assistance robot for elderly people with mild cognitive impairment. This aims to meet a growing need home help for the elderly living alone. Indeed, the aging population growing, it is estimated that 33% of French have more than 60 years in 2060. The challenge is to program an interactive robot (via its perceptual abilities) able to learn the relationship between the user and a subset of everyday objects of that or of the relevant objects, present or possibly lost in the shared space of the robot and the user. In this context, there are still many obstacles to be overcome, particularly in terms of: (i) joint perception of man and his environment, (ii) integration of a robotic system, (iii) validation scenarios involving the robot and an elderly person in interaction with some everyday objects. The project objective is to see the robot answer questions on ten common items (defined by a preliminary study on a population that complains of cognitive disorders) through appropriate actions.For example, the robot signal the location of an object by moving toward him, grasping or giving oral information as to its position if the object is not reachable. The RIDDLE multistakeholder project: it includes the Magellium society, Gérontopôle Toulouse, MINC team LAAS-CNRS and the company Aldebaran Robotics which the robot must serve as a platform for the final demonstrations. This thesis was jointly supervised by Frederic and Isabelle Lerasle Ferran ed faculty respectively in the RAP teams LAAS-CNRS and SAMoVA IRIT-UPS. During this project, we, in partnership with the Gérontopôle defined a robotic scenario involving three main phases. A monitoring phase when the robot is far away from the user and the observed position, waiting for a request of interaction, a phase of proximal interaction where the robot is closer to the user and interacts with it, and finally the transition that enables the robot to move from one phase to another. This scenario is constructed so as to create a robot proactive interaction but non-intrusive. The non-invasive nature is embodied by the monitoring phase. Proactivity is, in turn, evidenced by the creation of an e intentionality detector enabling the robot to understand non-verbal way the will of the user to communicate with him. The scientific contributions of this thesis cut across various aspects of the project: the robotic scenario, the intentionality detector, a filtering technique particle swarm, and finally a Bayesian technique improvement of the word error rate from ' distance information. This thesis is divided into four chapters. The first deals with the intentionality detector, the second of the developed filter technology, the third phase of the proximal interaction and techniques employed, and the last chapter focuses on robotic implementations.
Document type :
Complete list of metadata

Cited literature [94 references]  Display  Hide  Download
Contributor : Arlette Evrard <>
Submitted on : Thursday, September 8, 2016 - 10:02:29 AM
Last modification on : Thursday, June 10, 2021 - 3:06:01 AM
Long-term archiving on: : Friday, December 9, 2016 - 12:28:43 PM


  • HAL Id : tel-01291838, version 1


Christophe Mollaret. Perception multimodale de l’homme pour l’interaction Homme-Robot. Robotique [cs.RO]. Universite Toulouse III Paul Sabatier, 2015. Français. ⟨tel-01291838⟩



Record views


Files downloads