The goal of this four-semester student project is to develop a VR maze
and a set of artificial agents whose task it is to explore the maze,
either on their own or in competition. The agents will possess varying
object and scene recognition abilities which are shaped after
sensorimotor models of human object and scene recognition. Visual
information will be at the basis of the agents' knowledge acquisition.
Project subtasks include the design of suitable action-centered
knowledge representation structures and the development of exploration
strategies which operate on declarative, rule-based, or analogical
environmental knowledge and on uncertain knowledge on actions, among
others. The exploration scenario will be game-like.