Plenary Talk II, Wednesday May 17, 2006

Chair: Prof. Kevin Lynch, Northwestern University

1:30pm-2:30pm

Location: Palm Ballrooms 3, 4, and 5

Brain-Inspired Robotics: A Dynamical Systems Account For Cognitive Behavior

Dr. Jun Tani

Brain Science Institute, RIKEN

JAPAN

Abstract

While our understanding of the human brain is far from complete, novel neural mechanisms are being identified that may better inform efforts by robotics researchers confined by currently computational modeling approaches in designing machines. Such biological principles may open new routes to autonomous, learning robots. To illustrate, we know that patients with impairments in regions associated with higher visual processing cannot perceive objects consciously, yet they can grab or point to the object even if they are not consciously aware that the object there. It is known that this type of unattended behavior is generated in the dorsal path way of brains including the association cortex where the streams of the sensory perception and the motor generation are intermingled, thereby being inseparably processed. This contrasts with the conventional view that first the visual cortex recognizes target objects, then the motor cortex generates the corresponding action based on the information obtained.

Along with other emerging data, these findings indicate that the brain may not work through assemblies of localized functional modules (such as recognition, planning, action and learning) that regulate the entire system. Hence, current computational models may also be imprecise. We took a different brain-inspired approach and developed a system that assumes the presence of some type of global hermeneutic dynamics. These dynamics emerge from the sensory-motor coupling between the neuronal dynamics and physical movements of the body within an environment where each function appears distributed in its flow.

Our robotics research attempts to extract the hermeneutic dynamics of the brain and examine its essential mechanisms synthetically using a variety neuro-cognitive robotics experiments. These experiments include imitative interaction with human, learning object manipulation, and behavior-linguistic associative learning. Our robots successfully generated flexible and situated cognitive behaviors when hermeneutic dynamics were sufficiently self-organized following iterative learning in an environment. We also found that learning was more effective when the robot and a human shared direct experiences at the sensory-motor level. We would like to explore, here, the significant potential of such brain-inspired approaches to robotics to advance humanitarian robotics by providing a means to autonomy through a robot’s intrinsic properties and thereby facilitate closer interactions with humans.

Short Biography

Jun Tani received the B.S. degree in mechanical engineering from Waseda University, dual M.S. degrees in electrical engineering and mechanical engineering from University of Michigan and Dr. Eng. from Sophia University. He started his research career in Sony Computer Science Laboratory in 1990. He has been appointed as a team leader in Lab. for Behavior and Dynamic Cognition, Brain Science Institute, RIKEN in Tokyo since 2000 where he is currently leading 7 PostDocs. He was also appointed as a visiting associate professor in Univ. of Tokyo from 1997 to 2002.

He has studied the problems of robot learning with theoretical modeling of complex adaptive systems and neuronal networks for more than 15 years. He has been also interested in phenomenological problems of “self-consciousness” and his studies have addressed those problems from the view of embodied cognition. Five years ago, he started neuroscience studies on behavior learning processes in real brains utilizing both scheme of human brain imaging and animal electrophysiology. His envision is to establish the “brain-inspired robotics” by integrating these approaches.