Next-best-view planning for environment exploration and 3D model construction

Abstract

Smart robotics platforms in human environments usually develop a complete model of the surrounding environment. Exploring and modelling the environment required planning of best view positions as well as a navigation strategy that assure completeness and high quality of the construed model. In this paper, we describe a next-best-view planning for a mobile robot with limited sensors capability to explore and construct a 3D model of an indoor environment. Starting with a 2D occupancy grid map, the proposed algorithm shall process this map and generate candidates view poses. Next, the algorithm shall select the set of best views and generate a navigation plan of RGB-D sensor poses while satisfying a set of constrains in order to maximize the quality of the 3D constructed model and coverage of the indoor environment. Experimental results confirm the effectiveness and validity of the proposed system.