Skip to Main content Skip to Navigation
Theses

Endowing the Robot with the Abilities to Control and Evaluate its Contribution to a Human-Robot Joint Action

Amandine Mayima 1 
1 LAAS-RIS - Équipe Robotique et InteractionS
LAAS - Laboratoire d'analyse et d'architecture des systèmes
Abstract : Robots will interact more and more with humans in the future and thus will need to be endowed with the pertinent abilities. We are still far from having autonomous robots among humans and able to smoothly collaborate with them but the work of this thesis is a contribution bringing the community a bit closer to this goal.When humans collaborate to achieve a task together, numerous cognitive mechanisms come into play, more than we would have thought at first glance. Some of these mechanisms are also triggered in humans’ minds when they interact with robots as they are essential to a successful collaboration. Therefore, it is important for roboticists designing robots that will closely interact with humans to be aware of and take into account the humans mental states and sensorimotor functions involved in controlling and smoothing collaborative task performance. However, this does not imply that robots have to be endowed with the same mechanisms since being able to collaborate with humans does not mean to imitate them. What is key to roboticists is to understand how humans work and to design robots that will adapt.Consequently, this thesis starts with an immersion in philosophy and social and cognitive psychology. Then, we explore BDI and cognitive robotic architectures which have inspired us to design our own architecture in which, JAHRVIS — the main contribution of this thesis — endows a robot with the abilities not only to control, but also to evaluate its joint action with a human.Joint Action-based Human-aware supeRVISor (JAHRVIS) is what we call a supervision system, i.e., it embeds the robot high-level decisions, controls its behavior and tries to react to contingencies, always considering the human it is interacting with. It is able to do so by taking into account shared plans, human mental states, its knowledge about the current state of the environment, and human actions. JAHRVIS is designed in such a way that it is generic enough to handle various kinds of tasks.Not only JAHRVIS controls the robot contribution to a collaborative task, it also tries to evaluate if the interaction is going well or not. It is possible thanks to a set of metrics we have built and a method to aggregate them. We claim that having a robot with this ability allows it to enhance and make more pertinent its decision-making processes. In future work, this granularity will allow the robot to know precisely on what level it needs to act when a low Quality of Interaction (QoI) is assessed.JAHRVIS has been integrated in a cognitive robotic architecture and effectively deployed to achieve several collaborative and service tasks. These tasks demonstrated the robot’s abilities related to perspective-taking, planning, knowledge representation with theory of mind, manipulation, and communication.
Document type :
Theses
Complete list of metadata

https://hal.laas.fr/tel-03571963
Contributor : ABES STAR :  Contact
Submitted on : Monday, February 28, 2022 - 3:09:38 PM
Last modification on : Wednesday, June 1, 2022 - 5:03:52 AM

File

2021AmandineMAYIMA.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-03571963, version 2

Citation

Amandine Mayima. Endowing the Robot with the Abilities to Control and Evaluate its Contribution to a Human-Robot Joint Action. Automatic. INSA de Toulouse, 2021. English. ⟨NNT : 2021ISAT0008⟩. ⟨tel-03571963v2⟩

Share

Metrics

Record views

36

Files downloads

0