Recognition and movement in an artificial environment with a bipedal robot

Abstract

However well we control a walking bipedal robot, the images obtained by a camera are tilted to the left or right, and have small irregularities. This complicates the recognition of an environment by using a camera in a walking robot when the robot cannot move smoothly. The reason for using a bipedal robot is to make the robot as similar as possible to a human in body shape and behavior in order to make collaboration easier. This is difficult to attain with other types of robot such as wheel-driven robots (Sato et al. AROB2008; Fujiwara et al. WMSCI2009). In an artificial environment which mainly consists of vertical and horizontal lines, the tilt angle of camera images must be corrected by using the Hough transformation, which detects lines which are nearly vertical (Okutomi et al. 2004; Forsyth and Ponce 2007). As a result, the robot can successfully recognize the environment with stereo vision using images obtained by correcting the tilted ones.