A multiple camera based visual compass for a mobile robot in dynamic environment

Abstract

It is important to gather knowledge of orientation of the robot at any pose in many mobile robotic tasks. A visual compass is a technique that matches images to find the orientation of a robot. These images are collected at the reference and current pose of the robot. A robust visual compass technique is proposed to determine the orientation of the robot at any pose with respect to a reference direction using multiple cameras for unknown dynamic environment. The multiple camera system is the only sensor used to develop visual compass technique. The method estimates the orientation of the robot by finding first coarse and then fine angle difference between reference and current positions. The experiments in indoor and outdoor dynamic environment validate that the proposed visual compass technique estimates the orientation of a mobile robot with respect to its orientation at reference position within ±5° error. The proposed technique does not use expensive image processing tasks such as feature extraction and feature matching.