Long-term RFID SLAM using short-range sparse tags

Abstract

While on the path forward to the long-term or lifelong robotics, one of the most important capabilities is to have a reliable localization and mapping module. Data association and loop detection play critical roles in the localization and mapping problem. By utilizing the radio frequency identification (RFID) technology, these problems can be solved using the extended Kalman filter (EKF) based simultaneous localization and mapping (SLAM) with the tag information. But one of the critical barriers to the long-term SLAM is the overconfidence issue. In this paper, we focus on solving the overconfidence issue, which is introduced by the linearization errors. An Unit Circle Representation (UCR) is proposed to diminish the error in the prediction stage and a Correlation Coefficient Preserved Inflation (CCPI) is developed to recover the overconfidence issue in the update stage. Based on only odometry and sparse short-range RFID data, the proposed method is capable to compensate the linearization errors in both simulation and real experiments.