The development of corridor identification algorithm using omni-directional vision sensor

Abstract

Robot is one of applications in automation that has been implemented and developed in the last decade. The use of robots can be found in many manufacturing processes, and other activities like assembly, transportation, medical, and research. Based on its mobility, robot can be divided into two types: static robot (fixed at one place) and the mobile robot (can move from one point to another). One example of mobile robots is AMR (Autonomous Mobile Robot) which has a navigation system in guiding the robot to move towards the desired places. When an AMR placed on a corridor, it requires a system of sensors to provide feedback about the surrounding environment so that AMR can perform desired movements when moving in the corridor. In this study, the sensor used is omni-directional vision sensor. This sensor is capable of capturing images that provide information of its 360º surroundings in one image. This image requires further identification process in order to get the information contained in the image and use it as feedback for AMR. This identification process can be done by designing an algorithm to translate and produce decisions based on images captured by the sensor. The first step of this algorithm is to divide the 360º environment into four quadrants and then further information from each quadrant is used to determine the type of corridor. There are 4 types of cases that may occur in a quadrant as follows, straight road, dead ends, turn point, and T-junction. Based on the cases in quadrant, some conditions that may be found in the hallway of this research are blind alley, intersection, right turn, left turn, and other conditions. The designed algorithm is implemented in several images taken using the omni-directional vision sensors. Decision on conditions of corridor stated by the result of the algorithm is in accordance with the conditions of the corridor in the real situation. Weaknesses of the algorithm is the existence of specific constraints such as environmental contrast, the maximum distance of detection, etc.