2005 Fiscal Year Final Research Report Summary
Synthetic Study of Imitation on Humans and Intelligent Robots
| Project/Area Number |
13GS0006
|
|---|
| Research Category |
Grant-in-Aid for Creative Scientific Research
|
| Allocation Type | Single-year Grants |
|---|
| Research Institution | The University of Tokyo |
|---|
Principal Investigator |
TOMOMASA Sato The University of Tokyo, Graduate School of Information Science and Technology, Professor (50235371)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
INABA Masayuki Univ. of Tokyo, Graduate School of Information Science and Technology, Professor (50184726)
KUNIYOSHI Yasuo Univ. of Tokyo, Interfaculty Initiatives in Information Studies, Professor (10333444)
TADOKORO Satoshi Tohoku Univ., Graduate School of Information Sciences, Professor (40171730)
MORI Taketoshi Univ. of Tokyo, Interfaculty Initiatives in Information Studies, Associate Professor (20272586)
INAMURA Tetsunari Univ. of Tokyo, Graduate School of Information Science and Technology, Lecturer (20361545)
|
|---|
| Project Period (FY) |
2001 – 2005
|
| Keywords | Imitation Integrated System / Behavior Emergency / Causality Recognition / Intention Estimation / Body Image Acquisition & Adaptation / Learning from Observation / Imitation by shape / Behavior Imitation |
|---|
| Research Abstract |
Imitation is the most fundamental intelligence as for the integrated behavior system of the cognitive brain functions and the developmental ability to learn new behaviors from man to adapt real environment. The following researches were conducted aiming at the science and technology of imitation. The following results would form not only the nucleus to develop new intelligent systems but also theoretical basis of cognitive science research fields.
1) Construction of action observation and recognition systems :
An environmental embedded system to measure human motions and a smart and flexible data management framework are developed. Based on the observation system, daily behavior pattern discovery, extraction methods, and robust action recognition frameworks based on statistical machine learning techniques with massive human motion data are presented and verified.
2) Construction of action generation and correction systems :
Humanoid behavior modification methods via evolutional approaches
… More
and human teaching assistance approaches are developed. In the latter approach, humanoid recognizes situations of the real world and learn what should be paid attention.
3) Elucidation and modeling of human imitation function:
Building synthetic functional map on imitation : This map clarifies several aspects of behavior imitation by making correspondence to brain functions. A theory for whole-body dynamic action that is called global dynamics mechanism is proposed and verified by analyzing rolling and rising motion of human and humanoid. Novel self-organizing network models are constructed representing several functions of brain from low-level visual perception to high-level context processing. Experimental studies on emergence of intelligence and development of cognition of human baby is executed.
4) Experimental systems for realizing behavior imitation :
Behavior imitation systems are realized based on humanoid and room-typed intelligent robot systems in order to verify the techniques proposed in the above research achievement (1-3) works well. The result implies that intelligent systems with the proposed human action observation and recognition methods, behavior generation and modification methods are proved to achieve goal-level imitation tasks. Less
|
