Fog-Robotics Infrastructures Simulation-Based Sizing Approach

Nowadays, robots are increasingly being used to facilitate the automation of human tasks in various sectors and domains. These robots often have limited resources (memory, storage, computing power) for their activities (such as processing an AI service). In such cases, it is common to use on-demand architectures like Cloud architectures to offload the robot’s data management and optimize its operations (known as Robot-to-Cloud or R2C architectures). However, sizing such architectures involves solving highly complex optimization problems or comparing numerous hypothetical and costly scenarios in terms of time and/or money for deployment in a real environment. To address this challenge, we propose an approach called FRISSA (Fog-Robotics Infrastructure Simulation-based Sizing Approach) in this article. It estimates resource requirements in Cloud-based environments based on simulation results obtained with SimGrid. The approach specifically focuses on homogeneous and static Fog Robotics infrastructures. It introduces an algorithm that simultaneously optimizes the duration and cost of using Fog resources to determine the minimal resource requirements that compromise both metrics. We apply this approach to an existing robotics use case, which involves deploying a human face detection service on the Cloud for a stream of images.