Indoor exploration strategy for autonomous mobile robot based on direction map

The paper focuses on the variant of autonomous exploration problem that implies three-dimensional (3D) reconstruction of an unknown environment by moving a mobile robot with optical sensor through the environment. The authors introduce a data structure called direction map that accumulates information about viewpoints for every region of the scene. A direction map is used to formulate an exploration strategy according to the next-best-view principle. This strategy can be efficiently implemented using rectangular partition of an orthogonal polygon and integral matrices. Experimental evaluation of the proposed strategy is done in virtual environment. Completeness of the resulting 3D model is evaluated by comparison with the ground truth.