返回Google圖書搜尋

Acquisition of Obstacle Avoidance Actions with Free-Gait for Quadruped Robots

Tomohiro YamaguchiKeigo WatanabeKiyotaka Izumi

出版	INTECH Open Access Publisher, 2006
ISBN	386611284X9783866112841
URL	http://books.google.com.hk/books?id=hAyQoAEACAAJ&hl=&source=gbs_api

註釋We experimentally have proved a method for acquiring a path to a destination and obstacle avoidance of a quadruped robot. Robot actions were determined through an RBFNN, whose input consisted of destination information, obstacle configuration, and current robot status. Using training data on environmental conditions, focusing on x-, y-, and z-coordinates of different obstacles and certain destinations, RBFNN design parameters were optimized using a GA so that the robot reached the destination with a minimum number of walking cycles. For an untrained (unknown) environment, we found that the RBFNN was useful for acquiring an obstacle avoidance path to the destination. Effectiveness of this approach was examined by actual experiments. However, free-gait motion was not taken into consideration in the first reseach. A method of determining the order of swing leg in free gait by an RBFNN, whose inputs are the amount of movements for the quadruped robot and the height of the body, has been proposed for the second research. In the tuning of design parameters of the RBFNN, 20 data to which the amount of movements for the robot was changed are prepared for each order of swing leg. Such design parameters were optimized using GA so that the relation between an input and an output is satisfied.