Efficient exploration for real-time robot indoor 3D mapping

Abstract

Environmental mapping is an important problem for autonomous robots. In this paper, we study the problem of how to efficiently explore an unknown environment to build a 3D map. A Robotic 3D Mapping (RoM) system is developed which enables efficient exploration for real-time robot mapping. The 3D information comes from a RGB-D camera on the pan-tilt unit mounted on a Pioneer mobile robot base. In the mapping process, the camera pose is estimated based on a Bayesian framework combining the robot motion and visual features. The map is updated in real time and converted into the 3D occupancy map with new observed data. An efficient exploration strategy is proposed using the resulted 3D occupancy map. A multi-layer structure is applied to filter the viewpoints to a small set and the next best viewpoint is determined by maximizing the expected information gain. Experiments are conducted in an indoor environment and the results show that the robot is able to perform efficient, autonomous exploration to cover the unknown areas and build the map.