Covariance Visualisations for Simultaneous Localisation and Mapping

Abstract

Simultaneous Localisation and Mapping (SLAM) is a method of environment mapping in mobile robotics. One of the most popular classes of this algorithm is the Extended-Kalman Filter (EKF) SLAM, which maps the environment by estimating similarities between currently registered scene objects and newly perceived ones. More advanced versions of this algorithm are necessary, e.g. for multiple robots or outdoor environments. However, development is difficult because of the complex interaction between the internal robot state, the perceived scene and the actual scene. New visualisation methods are hence required to enable developers to debug and evaluate EKF-SLAM algorithms. We present novel Augmented Reality based visualisation techniques which display the algorithm’s progress by visualising feature and robot pose estimates, as well as correlations between features and clusters of features. The techniques allow a qualitative estimate of the algorithm’s mapping compared with the ground truth and indicate the correctness and convergence properties of the SLAM system. 1