Task learning of a task robot in real space by using a learning system in virtual space

Reinforced learning by which a robot acquires control rules by trial and error has attracted considerable attention. However, it is quite difficult for robots to acquire control rules by reinforcement learning in the real space because many learning trials are needed to arrive at the control rules; the robot itself may lose control, or there may be safety problems with the control objects. In this paper we propose a method in which a robot in the real space learns a virtual task, after which the task is transferred from the virtual to the real space. The robot eventually acquires the task in a real environment. We show that a real robot can acquire a task in a virtual space with an input device, using the example of an inverted pendulum. Next, we verify that the acquired task in the virtual space can be applied to a real‐world task. We emphasize the utilization of the virtual space to effectively obtain the real‐world task. © 2010 Wiley Periodicals, Inc. Electr Eng Jpn, 172(1): 38–47, 2010; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/eej.20968