Robot localization and mapping problem with unknown noise characteristics

Abstract

In this paper, we examine the H <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">∞</sub> Filter-based SLAM especially about its convergence properties. In contrast to Kalman Filter approach that considers zero mean gaussian noise, H <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">∞</sub> Filter is more robust and may provide sufficient solutions for SLAM in an environment with unknown statistical behavior. Due to this advantage, H <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">∞</sub> Filter is proposed in this paper, to efficiently estimate the robot and landmarks location under worst case situations. H <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">∞</sub> Filter requires the designer to appropriately choose the noise's covariance with respect to γ to obtain a desired outcome. We show some of the conditions to be satisfy in order to achieve better estimation results than Kalman Filter. From the experimental results, H <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">∞</sub> Filter performs better than Kalman Filter for a case of bigger robot initial uncertainties. Subsequently, this proved that <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">∞</sub> Filter can provide another available estimation method for especially in SLAM.