Online SLAM in dynamic environments

Abstract

In this paper, we propose a novel online algorithm for simultaneous localization and mapping (SLAM) in dynamic environments. We first formulate the problem with two interdependent parts: SLAM and multiple target tracking (MTT). To pursue online performance, we propose a hierarchical hybrid method to solve SLAM: locally by maximum likelihood (ML) with occupancy grid map, and globally by extended Kalman filter (EKF) with feature-based map. Meanwhile we apply a straightforward nearest neighborhood (NN) algorithm based on Euclidean metric to address MTT. In order to track multiple moving objects reliably, we propose an enhanced fuzzy clustering (EFC) method to segment 2D range images and reliably group objects. Experiments validated on Pioneer 2DX mobile robot with SICK LMS200 demonstrate the capability and robustness of the proposed algorithm