Simultaneous localization and map building using linear features

Abstract

In this paper the simultaneous localization and map build-ing (SLAM) problem, for a robot navigating in indoor envi-ronments, is addressed. A line-based representation of the environment is adopted. Line parameters are extracted from range scans, and the corresponding covariance matrices are computed from the statistical characteristics of the raw data. An Extended Kalman Filter is then used to simultaneously estimate the robot pose and update the line-based map. Fea-ture matching is enhanced by separately keeping track of the segment associated to each line. The proposed technique is validated through numerical simulations and experimental tests, featuring the mobile robot Pioneer 3AT within a real-world indoor environment. 1.