Multi-level occupancy grids for efficient representation of 3D indoor environments

Abstract

Mapping 3D environments is a fundamental yet challenging problem for mobile robot applications. Although 3D sensory data can be efficiently obtained using low-cost commercial RGB-D cameras, direct extension of the widely-adopted occupancy grids to 3D environments would cause problems, such as large storage consumption and intensive computation cost. In this paper, we propose to use a stack of 2D occupancy grids, each of which corresponds to a horizontal slice of the 3D environment at a specific height, as an efficient representation of 3D environments for indoor applications. Moreover, an existing algorithm based on Rao-Blackwellized Particle Filters (RBPF) is modified accordingly to perform simultaneous localization and mapping (SLAM) using the proposed multi-level occupancy grids (M-LOG), the entire codes of which have been made open source at https://github. com/AngelTianYu/micros mlog. Experimental results from both simulation and real-world tests validate the effectiveness of the proposed approach in indoor environments. Computational cost of the approach scales linearly with the number of 2D map slices, making it the user’s choice the trade-off between vertical map resolution and efficiency.