A context based decision making architecture for a navigation task

Souheïl Hanoune (1) , A. de Rengervé (1) , Pierre Andry (1) , Mathias Quoy (2) , Philippe Gaussier (1)

1 ETIS - UMR 8051 - Equipes Traitement de l'Information et Systèmes
2 Neurocybernetics

Corrective learning is an interesting paradigm for on-line learning of a trajectory from non-verbal interaction. We propose a model of action selection based on corrective interactions between the robot and a teacher. By using the appropriate correction signal, a minimal solution is to learn specific contexts inhibiting the wrong actions in order to let the appropriate behavior be exhibited in any circumstances. We present a solution to an action selection problem, where a mobile robot has to pick up objects in a given position of the environment.

Domaines: Neurosciences; Réseau de neurones [cs.NE]; Intelligence artificielle [cs.AI]; Robotique [cs.RO]

Souheïl Hanoune, A. de Rengervé, Pierre Andry, Mathias Quoy, Philippe Gaussier. A context based decision making architecture for a navigation task. Third International Symposium on Biology of Decision Making, May 2013, Paris, France. pp.25-26. ⟨hal-01063968⟩