A Fuzzy UV-disparity based approach for obstacles avoidance

Abstract

In this paper we present an approach which contributes to the avoidance of obstacles by a mobile robot during its navigation. This approach consists on the use of stereoscopic images in order to detect an eventual obstacle. For this, the used algorithm exploits the V-disparity image issued from the manipulation of the disparity map. After the detection of the obstacle, the robot estimates the position and dimensions of the obstacle in order to choose the suitable way to avoid it. The estimation of the obstacle properties is done by the treatment of the U-disparity image which is calculated in the same time as the V-disparity image. The avoidance approach is carried out using a fuzzy controller which generates commands to control the velocity of the different robot wheels so that the robot goes to the left or to the right of the obstacle.