ROSANA: Robot for Social Interaction in Unstructured Dynamic Environments

Abstract

Socially Interactive Robots must have the ability to interact with humans in a socially appropriate manner. Such a task requires a high level of perception of the robotic agent. A high degree of perception requires considerable computational power, something challenging to achieve in mobile robotics. Robots need to balance operating time, batteries (autonomy and weight), embedded sensors, and processing hardware. One way to overcome the mobile agents' perception limitations is to move the data processing capabilities to external processing units. In this article, we present the initial results of the ROSANA platform, a socially interactive robot for public environments capable of interacting with humans in the wild. ROSANA has a high level of visual perception, thanks to an architecture based on cloud services. We named this approach as Perception as a Service - PPaaS. Case studies exemplify the integration between the mobile agent and the cloud in the ROSANA architecture like detection, interaction with humans, and route planning considering humans in unstructured and dynamic environments.