Based on machine learning and bioelectrical signal,
we are studying how to understand and represent human motion and
how to deliver and train motion skills to robots.

Background and Objectives

  • Human motor skills differentiated from conventional robots in that they actively utilize the flexibility of the musculoskeletal system
  • Existing flexible robots standing out only in terms of interacting with objects or environments
  • Developing intelligent robots that reproduce or transcend
    • dynamic, elastic, and explosive human motor skills
    • motor skills expressing emotions in a sophisticated and delicate tasks


  • By merging motion information and muscle activities,
    • Estimate force or power generated by the muscle
    • Estimate mechanical impedance of joint or body segment
    • Predict motion intention before motion


  • To measure human motion
    • Estimate the use of muscle elasticity for explosive tasks
    • Estimate the interacting mechanical impedance in delicate tasks
  • To discriminate and evaluate human motion
    • Organize and express motion path in an implicit and hierarchical way
    • Express and generate motion with mechanical impedance
    • Analyze motion data to understand human motion principles
    • Find out what differences occur physically according to the skill level
  • To reproduce or transcend human motion by a robot
    • Develop a robot system that implements dynamic/explosive movement of humans or delicate and flexible movements
    • Develop impedance robot control algorithm overcoming motion differences

Last Updated on September 15, 2020