Social mapping on RGB-D scenes

The more the robotics technology tends to get established in human everyday life, the greater the necessity for robots to obtain social skills, facilitating their presence and behavior among humans. In order for robots to attain socially-aware characteristics, their navigation strategy should comply with a set of certain criteria, such as the respect of spatial interactions. Towards this end, the paper in hand presents a framework that operates on a single RGB-D scene. Initially it seeks the human presence and defines respective bounding boxes. Then, within the 3D reconstructed scene the people in the bounding boxes are replaced by points following a Gaussian distribution. The latter results to a grading occupancy grid, which complies with the spatial human comfort zones.