Correcting Odometry Errors for Mobile Robots Using Image Processing

摘要

Abstract- The mobile robots that are moving in partially known environments have a low availability, the mean time between failures being around 17.5 hours. From the total number of failures, 25 % are localization failures caused mainly by odometry errors. Without a proper odometry error correction procedure, a wheeled mobile robot is not able to determine correctly its position and therefore it is lost. This paper proposes a solution to obtain fault tolerance to odometry errors (both systematic and nonsystematic errors) using an online correction method based on image processing. During the movement, if odometry errors are occurring, the orientation of the mobile robot will be corrected using information taken over from certain points of the environment (visual reference points) through images. The influence of the proposed correction method in the overall system’s availability will be shown.