Symbiotic Intelligence group focuses on high-level cognitive functions, and takes more synthetic approach to understand basic principles of intelligence reformulated in the engineering context. Consider how human learned to fly. Leonard da-Vinci, for example, tries to mimic wings and flipping motion of birds, and failed. The Wright Brothers discovered that the essence of flight is "lift" and "thrust", and created a fixed wing to generate lift when thrust is applied using an engine. The recent success of the IBM's Deep Blue that beat human chess champion is decomposition of human intelligent ability during the chess game into memory and search. While the demonstrated success is limited to the chess, it may well be applicable for other highly intelligence tasks which shares essential aspects with chess.

Symbiotic Intelligence group strive to find the underlying principles of intelligence using high degree of freedom robot systems with multiple perception channels, such as vision, auditory, tactile, etc. We consider the integration of multi-modal perceptive inputs combined with motor control to effects the environment and self status is the basic conditions that exhibits intelligence. The concept of streams of multiple perception channels and their association is the other important premises. We have not clearly decomposed that are essence in the engineering reformulation, as identification of such concept is the final goal of the project.

Specifically, two scientific projects and one design initiative are currently being undertaken.

Humanoid Project

First, Humanoid project uses a human-like robot with high degree-of-freedom and multiple sensor inputs. Current platform implements upper body, and have five DOFs for head and pan-tilt for both of two eyes. While current version does not have arms and legs, we are planning to design and develop a whole body in the next step. At the initial stage of the project, we have implemented an integrated system of auditory scene analysis and visual processing to bootstrap processing capability of both perception. This result will be reported in "Using Vision to Improve Sound Source Separation," Proc. of AAAI-99 (to appear). In the next step, we intent to extend the notion of stream integration to incorporate tactile sensing.

Real-time auditory and visual tracking implemented on the upper-torso humanoid, SIG ( Finished FY2003 )

  • Sound Source Separation by Integrated Processing of Audio and Vision (Below chart)

Biped walking robot, PINO ( Finished FY2001 )

Small-sized Humanoid robot, morph ( Finished FY2003 )

photo by Yukio Shimizu

Robot Design Initiative ( Finished FY2001 )

While traditional robotics pursue functions and capabilities of robots, as technology matures, "design" emerges as a critical factor in robots. This is particularly true for robots for consumers, such as entertainment robots, home assistant robots, and other robots that are expected to be seem by general people, instead of factory robots that are confined inside factories. In order to facilitate the progress in the field for the era of "robots@home", we propose Robot Design as a new field in the industrial design.

  • Robot Design Initiative
  • Awards
    • MITI Good Design Award (2000)
    • Venetian Biennale Invitated Exhibition (2000)
    • Prix Ars Electronica (Austria, 2000)
    • Museum of Modern Art (MoMA) Invited Exhibition (2001)

  • IRoDA (=International Robot Design Association)

RoboCup J-STAR Project ( Finished FY1999 )

The second area of research focuses on multiple mobile robots, and how their sensors can be integrated and exhibits teamwork under dynamic environment. J-Star99 participating in RoboCup is the team of the project. We also intent to develop modular and reliable robot systems that can be a basis of mobile robot research including RoboCup.

Copyright (C) 1999-2003 Kitano Symbiotic Systems Project All Rights Reserved.