Cloud, fog, and dew robotics: architectures for next generation applications

Robots are rapidly moving from isolated to connected systems, with more and more operations offloaded to external systems. Cloud Robotics has indeed become a new important trend in this research field. In this paper we advocate the necessity to move a step forward in this direction considering the real needs of new generation robotics systems. Such systems have indeed very peculiar requirements with respect to the traditional applications that have progressively exploited Cloud infrastructures. Cloud is surely an important addition, but robots still require important computations to be performed closer to or onto them for, e.g., critical decisions involving human interaction. This motivates our work in which we delve into the possible alternatives to pure Cloud Robotics, considering Fog- and Dew-computing as better suited to next generation applications. We describe these complementary but alternative architectures and how they match the requirements of robotics applications. We identify Dew Robotics as the most promising architecture and motivate our choice. We present three use cases we have been working on, to better illustrate our proposal. We finally draw some conclusions for this emerging yet important research field.