Task-Model Based Human Robot Cooperation Using Vision
We are currently engaged in advancing research in human-robot cooperation with the goal of realization of the service robot in homes and offices. In order to assist humans, robots must correctly identify the tasks that the humans intend to perform, and then devise strategies to perform those tasks. To accomplish this, we first define the framework of the robot's task model for assisting the human's work. The robot then automatically generates a task model from event analysis based on observation of a human performing the desired task. Based on that task model, the robot plans appropriate assistance that it executes at the time of its cooperative work with the human. We have tested he effectiveness of this method by performing an experiment in which a human and a robot cooperated in assembling toy parts.
Publications
[1] H.Kimura and K.Ikeuchi: Vision Based Cooperation between Human and Robot, Robotics Symposia 97, pp.120-126, 1997.[2] H.Kimura, T.Horiuchi and K.Ikeuchi: Task-Model Based Human Robot Cooperation Using Vision, Proc. on Inter. Conf. on Intelligent Robots and Systems, pp.701-706, 1999.
[3] Y.Sato, H.Kimura and K.Ikeuchi: Extraction of functions of objects through observation of human demonstrations, Robotics and Mechatronics '01, 2A1-C12, pp.43, 2001.
[4] Y.Sato, H.Kimura and K.Ikeuchi: Description and implementation of robot operation based on the object functions, Abs. of the 19th Annual Conf. of the Robotics Society of Japan, pp.108, 1K26, 2001.
[5] Y.Sato, H.Kimura and K.Ikeuchi: Task analysis based on observing hands and objects by vision, 7th Robotics Symposia, pp.125-130, 2002.
Fig. Experiment results of the human-robot cooperation.
There are three types of robot assistance:
(A) Assistance to realize the purpose of a task
(B) Assistance for working efficiently
(C) Assistance by passing an object to the human