An Adaptive Motion Learning Architecture for Mobile Robots
Byeong-Kyu Ahn, Seung Min Baek, Tae‐Yong Kuc · 2006 SICE-ICASE International Joint Conference · 2006
Many studies are being conducted for the design of autonomous mobile robots which can learn adaptive motions without specific prior knowledge in dynamically changing environments. To this end, it is important to learn general motion rules, while excluding the assumptions limited to specific environments. In this paper, an adaptive motion learning architecture is presented. The architecture enables robots to avoid obstacles in unexperienced environments using the previously entered basic motions and generate generalized motion rules using the learned information. To collect learning data, robot conducts wandering motions to obtain and transmit the sensor information, distance to obstacles, and collision information to the learning architecture, which generates the motion of robot using the basic and learned adaptive motion rules. Lastly, the applicability is verified by learning obstacle avoiding motion in virtual space similar to actual space, by simulation