Exploration and observation planning for 3D indoor mapping

This paper deals with an observation planning for indoor 3D mapping. We consider the case where the robot makes a map with different resolutions; it observes informative regions from near positions. Since observing a region with a higher resolution requires more time, we need to carefully choose viewpoints for such observations. This paper therefore focuses on viewpoint planning for observing a set of informative (or important) regions. We develop a robot system which first explores an environment to make a 2D map using a 2D LIDAR and then uses the map for localizing informative regions and plans a sequence of viewpoints from which all of the regions can be observed with a RGB-D camera. The viewpoint planner tries to minimize the sum of view and travelling cost. We also investigate the trade-off between planning cost and plan quality.