Skip to Main content Skip to Navigation
Theses

Modélisation des mécanismes émotionnels pour un robot autonome : perspective développementale et sociale

Cyril Hasson 1
1 NEURO
ETIS - UMR 8051 - Equipes Traitement de l'Information et Systèmes
Abstract : The objective of this thesis is to draw inspiration from the neurobiology to model low level emotional mechanisms on a robot evolving in real environment. This work presents an emotional model coherent with experimental data describing the functioning of the cerebral structures involved in emotional mechanisms. Emotions play a central part in the regulation of behavior of humans as well as animals. In agreement with the darwinian view, emotions are seen as adaptive mechanisms enhancing survival. However, their organization around essential positive and negative signals gives them a dimensional flavor. Our model considers emotions as the result of the interaction dynamics between two systems. These systems allow the evaluation of the interactions with the physical and the social environment. This bio-inspired approach of emotions gives robots a basic framework to construct their behavioral autonomy and their communication skills. In this thesis, we show that they allow the robot to adapt itself to the characteristics of the environment as well as they underlie non verbal communication. The bio-mimetic approach of this thesis is reflected in methodological terms by the use of artificial neural networks for robot control architectures but also in functional terms by the organization of these networks as models of different brain structures and their interactions (amygdala, accumbens, periaqueductal grey, hippocampus, prefrontal cortex). Following the animat paradigm, the robot is seen as an animal which vital needs are satisfied by the resources of the environment. Experimentation are conducted on navigation behaviors relying on visuo-motor conditionings (visual strategy) and on path integration (proprioceptive strategy). Conditionings between nociceptive or hedonic signals and other sensory information or actions of the robot are at the basis of emotional regulation. The robot predictions allow it to learn aversive or appetitive behavior in response to its "pain" or "pleasure" expectations. The robot can also monitor its predictions to assess the effectiveness of its behaviors. This enables it to regulate its motivations and select its strategies (visual navigation or proprioceptive) and goals (environmental resources) in order to best meet its internal balance depending on its environment. This use of low level positive and negative signals allows to build a minimal emotional model providing autonomy to the robot behavior. In a second step, we use the emotional expressiveness as the basis for communication with the robot. A mechanical head enables it to express its emotions through its facial expressions. This communication consists in giving the robot reward and punishment signals. This exchange of information with the robot allows it to learn to valuate its environment or its behavior and thus to learn interactively to solve its navigation tasks. The model of emotional mechanisms presented in this work allows to investigate issues of autonomous robotics as well as issues of Human-Robot interactions. Moreover, this approach shows the interest of putting robotics at the heart of cognitive sciences due to the perspective given by the analysis of robot's behaviors supported by relatively simple neuronal architectures.
Document type :
Theses
Complete list of metadatas

Cited literature [262 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00904481
Contributor : Michel Jordan <>
Submitted on : Thursday, November 14, 2013 - 3:32:40 PM
Last modification on : Thursday, December 17, 2020 - 12:14:32 PM
Long-term archiving on: : Saturday, February 15, 2014 - 4:35:32 AM

Identifiers

  • HAL Id : tel-00904481, version 1

Citation

Cyril Hasson. Modélisation des mécanismes émotionnels pour un robot autonome : perspective développementale et sociale. Informatique. Université de Cergy Pontoise, 2011. Français. ⟨tel-00904481⟩

Share

Metrics

Record views

742

Files downloads

3018