Mobile Robot Localization Failure Detection

Localization failure detection is essential for realizing truly autonomous robot systems. In this paper, we propose a novel method that can resolve this problem. We employ a hypothesis testing to model the decision process that distinguishes whether the localization has failed or not, and propose a statistical model for the detection of localization failures. Through evaluations on publicly available datasets and actual-robot-platform, we demonstrate the efficacy on the proposed method. Experimental results show that the proposed method can perform robust localization failures detection in challenging scenarios.