CloudRobotics - Distributed Robotics using Cloud Computing

Book excerpt: Cloud Computing is a paradigm shift in computation that has been gaining traction over the recent years, which is supported by the increasing availability and ubiquity of a reliable wireless connection to the Internet. Cloud Computing enables the access to seemingly unlimited computer resources that are located on an external computer cluster (the Cloud). In contrast, some robots, e.g. drones, have mobility requirements such as maximum size/weight or minimum autonomy, and carrying more onboard computer resources usually means hindering these requirements. This principle can be brought to the field of Robotics hence the name Cloud Robotics. In this case, the goal is to allow robots to perform tasks they would not be able to under normal circumstances and/or to free onboard resources so that more tasks or more complex tasks can be executed at the same time by a mobile robot. There are many existing robotic tasks that can take advantage of massive processing power and storage, such as simultaneous localization and mapping (SLAM), navigation and trajectory planning, image processing, pattern recognition, human-robot interaction and machine learning to name a few. All of these can quickly drain the robot out of its computer resources, especially if some of these tasks are running at the same time. However, in order to access and export data to the Cloud some bandwidth is needed, thus making the system a tradeoff: on the one hand, computation load and storage space is being freed, while on the other hand more strain is being put on the wireless network usage. As wireless connection protocols become more and more powerful, a Cloud-based solution becomes more interesting. This dissertation aims to analyse this tradeoff by adapting two existing multi-robot tasks, working on the Robotic Operating System (ROS), and compare the Cloud-based approach to the traditional one. To validate the capabilities of Cloud-based robotic systems, both simulations and experiments with real robots were conducted. Simulation results show a clear gain in CPU time, while the latter confirms the outcome of the tasks remains the same. Despite the Cloud-based systems, requiring considerably more bandwidth, a modern off-the-shelf Wi-Fi router can provide with enough to support any realistic team of robots.