A sparse-dense approach for efficient grid mapping
Eurico Pedrosa, Artur Pereira, Nuno Lau · 2018
The regular volumetric grid is a popular method used in mapping to represent the environment, however for large three-dimensional environments it requires a large amount of memory that may not even be available. In this paper we present a sparse-dense data structure to manage the space of a volumetric grid that provides improvements over the octree data structure, a data structure popularized by the OctoMap mapping framework. Furthermore, we propose an online data compression scheme supported by a cache mechanism that further improves the space efficiency of our approach without compromising time efficiency. The approach is evaluated using public available datasets that show an increase in space and memory efficiency over OctoMap without compromising accuracy.