With the development of robotics, the use of autonomous mobile robots in the execution of everyday tasks tends to become increasingly feasible in different parts of society. This work aims to propose a graph based algorithm to identify possible associations of individuals in the environment, presenting a new approach to compute the level of interaction between individuals according to their position and body orientation. Subsequently, based on this detection of the interaction groups we propose a representation of the collective social space based on a density function through the junction of asymmetric Gaussian functions to delimit the socially acceptable space, so that a mobile robot can perform tasks in the presence of pedestrians without offering any type of physical risk or social discomfort like colliding with an individual or invading your personal interaction space. The proposed methodology for determining the association of individuals was evaluated using synthetic and real-world datasets, outperforming the state-of-the-art methods on most cases. In addition, we evaluated through simulations the representation of the collective social space demonstrating that our approach generates more socially favorable behaviors than methods that do not consider such restrictions.

Prof. Douglas Guimaraes Macharet – Advisor (DCC – UFMG)
Prof. Mario Fernando Montenegro Campos (DCC – UFMG)
Prof. Luiz Chaimowicz (DCC – UFMG)
Prof. Armando Alves Neto (DEE – UFMG)