1. Title: [KIDCS] Computational model of neuro-musculo-skeletal system: human biped walking
2. Date and Time: Thursday, April 24, 2008, 16:00 ~ 18:00
3. Place: E1 Seminar Room, Mechanical Engineering Building (N7)
4. Speaker: Prof. Sungho Jo (Div. of Computer Science, KAIST)
Ph.D. in electrical engineering and computer science from MIT (2006)
Research interests include intelligent robots, human-robot interaction,
computational intelligence, computational neuroengineering,
biomimetic system design, bio-inspired learning and control, etc.
The neurobiological system modeling may be motivated in two perspectives. First, the modeling could help understand the mechanismsor principles of biological system and, secondly, provide useful ideas about the development of robots or artificial systems. In this seminar, a neuro-musculo-skeletal model of human biped walking is introduced, and interesting ideas provoked throughout the model are discussed. Humanoid robotic locomotions or prosthetic legs are exemplified to discuss the usefulness of neurobiological system research.
The proposed model is built by integration of explicit models of neural systems. The whole system of human biped walking consists of cerebro-cerebellar interaction, spinal pattern generator, spinal synergies, peripheral reflex, and spring-like muscular actuator. A role of each neural system is proposed: The suprasegmental (cerebro-cerebellar) system controls postural balance by constructing feedback loop while spinal nervous system generates walking patterns feedforwardly. Spinal synergies facilitate the control of overall muscular behaviors. Each neural system model is discussed neurophysiologically. The integrated model is evaluated in comparison with human in terms of kinematics, dynamics, interaction with the ground, muscular activities, robustness, stability etc. In addition, simple behaviors modulated from nominal walking are realized. The proposed model demonstrates some features as follows. 1) The interaction between neural systems produces a desired motion. 2) Functionally decoupled neural structure enables modulated behaviors by simple local adjustment. 3) Dynamic motion generation may be possible even without detailed inverse dynamics computation or model prediction.
6. The forthcoming seminar schedule is as follows
Division of Electrical Engineering (E3-2)
KI for IT Convergence
Development of future city based in construction-IT convergence technology
Prof. Kwak, Hyo-Gyoung
Biomedical Research Center (E7)
KI for BioCentury
Peptide Self-Assembly: From Neurodegenerative Diseases To Novel Nanomaterials
Prof. Park, Chan Beum