Fusing odometry and sparse UWB radar measurements for indoor slam

Abstract

For security applications and in situations where optical sensors are not working, ultra-wideband (UWB) radar is an alternative technology for localization, mapping and object recognition. This paper presents an approach for solving the simultaneous localization and mapping (SLAM) problem for an autonomous robot with a small UWB radar array. Feature-based mapping in conjunction with an underlying state space model enables the reconstruction of the room with accuracy up to 10 cm. Two different ways of dealing with the data association problem — the task of sorting the measured time-of-flight values — are presented. Data fusion with odometry information is proposed to reduce the number of measurement steps. Experimental results with an autonomous robot show the feasibility of the concept.