Experimental study on using visual odometry for navigation in outdoor GPS-denied environments

Abstract

This paper presents an experimental study on utilizing visual odometry (VO) for pose estimation of mobile robots in outdoor GPS-denied environments. The experiments have been done using two different stereo VO algorithms implemented in Robot Operating System (ROS). The results have been compared with ground truth and drift has been identified. Sensor integration has been successfully used to minimize the drift. Challenges and possible solutions have been presented for future research and development.