Raisonnement sur le contexte et les croyances pour l’interaction homme-robot

Grégoire Milliez 1
1 LAAS-RIS - Équipe Robotique et InteractionS
LAAS - Laboratoire d'analyse et d'architecture des systèmes [Toulouse]
Abstract : The first robots appeared in factories, in the form of programmable controllers. These first robotic forms usually had a very limited number of sensors and simply repeated a small set of sequences of motions and actions. Nowadays, more and more robots have to interact or cooperate with humans, whether at the workplace with teammate robots or at home with assistance robots. Introducing a robot in a human environment raises many challenges. Indeed, to evolve in the same environment as humans, and to understand this environment, the robot must be equipped with appropriate cognitive abilities. Beyond understanding the physical environment, the robot must be able to reason about human partners in order to work with them or serve them best. When the robot interacts with humans, the fulfillment of the task is not a sufficient criterion to quantify the quality of the interaction. Indeed, as the human is a social being, it is important that the robot can have reasoning mechanisms allowing it to assess the mental state of the human to improve his understanding and efficiency, but also to exhibit social behaviors in order to be accepted and to ensure the comfort of the human. In this manuscript, we first present a generic framework (independent of the robotic platform and sensors used) to build and maintain a representation of the state of the world by using the aggregation of data entry and hypotheses on the environment. This infrastructure is also in charge of assessing the situation. Using the state of the world it maintains, the system is able to utilize various spatiotemporal reasoning to assess the situation of the environment and the situation of the present agents (humans and robots). This allows the creation and maintenance of a symbolic representation of the state of the world and to keep awareness of each agent status. Second, to go further in understanding the situation of the humans, we will explain how we designed our robot with the capacity known in developmental and cognitive psychology as "theory of mind", embodied here by mechanisms allowing the system to reason by putting itself in the human situation, that is to be equipped with "perspective-taking" ability. Later we will explain how the assessment of the situation enables a situated dialogue with the human, and how the ability to explicitly manage conflicting beliefs can improve the quality of interaction and understanding of the human by the robot. We will also show how knowledge of the situation and the perspective taking ability allows proper recognition of human intentions and how we enhanced the robot with proactive behaviors to help the human. Finally, we present a study where a system maintains a human model of knowledge on various tasks to improve the management of the interaction during the interactive development and fulfillment of a shared plan.
Document type :
Theses
Automatique / Robotique. Institut National Polytechnique de Toulouse - INPT, 2016. Français
Liste complète des métadonnées

Cited literature [116 references]  Display  Hide  Download

https://hal.laas.fr/tel-01591864
Contributor : Christine Fourcade <>
Submitted on : Friday, September 22, 2017 - 10:22:03 AM
Last modification on : Thursday, January 11, 2018 - 6:26:55 AM

Identifiers

  • HAL Id : tel-01591864, version 1

Citation

Grégoire Milliez. Raisonnement sur le contexte et les croyances pour l’interaction homme-robot. Automatique / Robotique. Institut National Polytechnique de Toulouse - INPT, 2016. Français. 〈tel-01591864〉

Share

Metrics

Record views

60

Files downloads

33