Follow me: interleaving human tracking and interacting with a new assistant robot

The development of socially-aware autonomous robots for being part of our daily lives is a pressure matter. Although almost all current robots use multiple sensors, the limitations of these autonomous systems to coexist with people are still evident: they have difficulties fusing all this information in an agile and consistent way and thus understand human intentions and act accordingly. The main consequence of these limitations is that today’s robots do not respond adequately to these human behaviours and, for example, result in forced and unnatural navigation. Today’s robots strive to have complete knowledge of the surrounding environment to make the right decisions at the right time. However, this purpose requires a perspective that integrates sensors of different natures and, based on it, a multi-modal perception of actuation. This paper presents a multi-modal and distributed architecture for the perception of the environment that uses information given by the robot sensors. The results of the proposed architecture are validated in the Follow Me use a case in a real environment.