Systèmes de navigation biologiques et artificiels : apprentissage de routes par des fourmis et par un robot mobile autonome
Résumé
Specification abstract
A major aim of mobile robotic is to provide machines with a greater autonomy allowing them to move about in and manipulate their environment independently of human intervention. In the living world, several insect species have developed efficient and robust navigational abilities in the course of evolution. This has motivated roboticists to look for biological navigation mechanisms that can be implemented on an autonomous mobile robot. This spatial ability is particularly astonishing in social insects (ants, bees, wasps), so-called central place foragers, which use different set of navigational strategies to reach back their home after each foraging excursion despite rather simple neurosensory equipment. Our project, conducted in close collaboration between researchers working in the field of informatic/robotic and ethology, was first to develop locomotion algorithms directly inferred from the spatial behaviour of two ant species (Cataglyphis cursor & Gigantiops destructor) studied both in the field (semi-desert and tropical terrains) and in visually controlled laboratory conditions. The algorithms were then successfully implemented in a biomimetic robot prototype moving from place to place on the basis of the ants' navigational strategies. Extension of this preliminary work on mass-produced robot models should require more logistical and financial investments.
A major aim of mobile robotic is to provide machines with a greater autonomy allowing them to move about in and manipulate their environment independently of human intervention. In the living world, several insect species have developed efficient and robust navigational abilities in the course of evolution. This has motivated roboticists to look for biological navigation mechanisms that can be implemented on an autonomous mobile robot. This spatial ability is particularly astonishing in social insects (ants, bees, wasps), so-called central place foragers, which use different set of navigational strategies to reach back their home after each foraging excursion despite rather simple neurosensory equipment. Our project, conducted in close collaboration between researchers working in the field of informatic/robotic and ethology, was first to develop locomotion algorithms directly inferred from the spatial behaviour of two ant species (Cataglyphis cursor & Gigantiops destructor) studied both in the field (semi-desert and tropical terrains) and in visually controlled laboratory conditions. The algorithms were then successfully implemented in a biomimetic robot prototype moving from place to place on the basis of the ants' navigational strategies. Extension of this preliminary work on mass-produced robot models should require more logistical and financial investments.
Donner la capacité à des machines de se déplacer de manière autonome est l'un des objectifs poursuivi par la robotique mobile. Cette forme d'autonomie par rapport à l'environnement, fondée principalement sur la motilité associée à des capacités d'orientation spatiale, est particulièrement développée au sein du monde vivant chez certaines espèces d'hyménoptères (fourmis, abeilles, guêpes) capables de résoudre des problèmes de navigation complexes en dépit d'un système nerveux relativement simple. Les caractères de fiabilité, de robustesse et de performance de ces mécanismes de navigation chez l'insecte, sélectionnés au cours de l'évolution biologique, peuvent alors servir de modèle pour l'élaboration de systèmes artificiels de navigation autonomes d'inspiration biomimétique. Notre projet, mené en synergie entre informaticiens/roboticiens et éthologistes, avait pour ambition l'élaboration d'algorithmes de déplacements, directement inspirés des mécanismes de navigation de deux espèces de fourmis (Cataglyphis cursor et Gigantiops destructor) que nous étudions au laboratoire et dans leur milieu naturel, et leur implémentation sur un robot mobile autonome. Ce travail a été concrétisé avec succès sur un prototype de robot-fourmi pourvu des règles de navigation directement inférées du comportement individuel de ces deux espèces de fourmis qui vivent dans des milieux assez hostiles (respectivement, milieux semi-désertique et tropical). L'extension de ces travaux sur des robots de série programmables et leur valorisation éventuelle sur des robots de type industriel nécessiteraient une logistique et des financements plus conséquents.