Skill Modeling and Robot Teaching
(Prof. T. ARAI and Y. MAEDA)

Cooperative Manipulation between Human and Robot with Motion Estimation (MPEG-1 Movie)
Robotic Manipulation by Human Demonstration and Motion Planning
Human Demonstration, Playback of Demonstrated Manipulation (MPEG-1 Movies)

Today’s robots demand detailed and complicated motion teaching to make them work. In order to enable the robots to perform various tasks flexibly, we have to model “skills” for robot tasks and realize skill-level robot teaching. Currently, we are studying skill representation for robots and skill transfer for several tasks as follows:

- Human-Robot Cooperative Rope Turning Based on Rhythm Entrainment (Fig. 1)

- Human-Robot Cooperative Manipulation with Motion Estimation using Minimum-Jerk Model (Fig. 2)

- Analysis of Human “Kendama” Motion by Non-Contact Impedance and Hidden Markov Models (Fig. 3)

- Teaching of Grasp/Graspless Manipulation by Human Demonstration and Motion Planning (Fig. 4)

Mathematical modeling of human skills and/or robot control inspired by human dexterity is conducted on the above robot tasks. We hope these research leads to the development of novel methodology for robot teaching.

Keywords: Robot Teaching, Skill Representation, Skill Transfer

References

Yusuke MAEDA, Atsushi TAKAHASHI, Takayuki HARA and Tamio ARAI: “Human-Robot Cooperation with Mechanical Interaction Based on Rhythm Entrainment ––Realization of Cooperative Rope Turning––,” Proc. of 2001 IEEE Int. Conf. on Robotics and Automation, pp. 3477~3482, 2001.

Yusuke MAEDA, Takayuki HARA and Tamio ARAI: “Human-Robot Cooperative Manipulation with Motion Estimation,” Proc. of 2001 IEEE/RSJ Int. Conf. on Intelligent Robots and Systems, pp. 2240-2245, 2001.

Fig. 1 Human-Robot Cooperative Rope Turning Fig. 2 Human-Robot Cooperative Part Handling

Fig. 3 Non-Contact Impedance Model Fig. 4 Manipulation Teaching from Human Demonstration
for Kendama