Friday Keynote: The Brain-like Vision

Chair:

Prof. Dr. Hans Meuer, ISC’09 General Chair, Prometeus & University of Mannheim, Germany

Keynoter:

Prof. Dr. Edgar Körner, President, Honda Research Institute Europe, Germany

Intelligence is a technology and a strategy for robust and flexible problem solving in complex environments (both natural and artificial) under the constraints of limited resources (e.g. time, energy). The need for intelligence becomes particularly visible when dealing with humanoid robots which are expected to behave like humans and which are in reality still much closer to their ancestors at the assembly line. Understanding essential principles of how the brain organizes behavior may enable us to provide our technical artifacts at least with some aspects of brain-like intelligence. Our approach is based on the assumption that the essence of brain computing is not in the local processing or learning algorithm but in the way the brain organizes processing. The challenge of such an approach is that rather than modeling isolated subsystems, large-scale computational models of complete functional blocks at several interacting levels of complexity have to be investigated. The simulation of large-scale hypotheses on brain function is limited by the available technology. Therefore, the Honda Research Institute Europe, we investigate different levels in parallel and convey fundamental results between these levels in order to circumvent the incorporation of all complexity levels in one system set-up. We target control architectures that are required for brain processing at the following levels: (1) the control of growth processes and development by gene-regulatory networks; (2) the detailed cortical columnar architectures for self-referential control for storing experience; (3) the behavior based dynamic allocation of systems resources for predictive visual scene analysis; and (4) the global behavior control architecture for autonomous interaction of our humanoid robot Asimo. Following the short outline of the basic philosophy of our approach, the focus of the talk is on above mentioned levels (3) and (4). Step by step we implemented nested control loops for reflexes, attention modulated behavior, on-line learning from sensory experience, and prediction/expectation driven behavior. This enables Asimo to learn to recognize objects through the interaction with humans, to learn associations between acoustic and visual objects, as well as to associate sound with behavioral concepts and to demonstrate simple prediction driven behavior.