A computer vision-aided motion sensing algorithm for mobile robot's indoor navigation

Abstract

This paper presents the design and analysis of a computer vision-aided motion sensing algorithm for wheeled mobile robot's indoor navigation. The algorithm is realized using two vision cameras attached on a wheeled mobile robot. The first camera is positioned at front-looking direction while the second camera is positioned at downward-looking direction. An algorithm is developed to process the images acquired from the cameras to yield the mobile robot's positions and orientations. The proposed algorithm is implemented on a wheeled mobile robot for real-world effectiveness testing. Results are compared and shown the accuracy of the proposed algorithm. At the end of the paper, an artificial landmark approach is introduced to improve the navigation efficiency. Future work involved implementing the proposed artificial landmark for indoor navigation applications with minimized accumulated errors.