A Platform for Cloud Robotics

This paper presents the evolution of a software platform for supporting experimentation in mobile robotics as part of teaching and researching activities. Starting with Web-based laboratories (WebLabs) in the early 2000s the platform kept evolving according to the networking and distributed computing trends since then. In addition to the physical resources managed by the platform, the platform now is able to manage a pool of virtual machines as resources for experimentation. This new class of resources brings the processing power as required by many modern mobile robotics applications. Virtual machines can be widespread on a cluster of processors, on a private cloud computing infrastructure, or on a public cloud computing service. Like any other resource managed by the platform the access to the virtual machines is subjected to user authentication and authorization. A mechanism of user authentication and authorization based on federated identities (single-sign-on) allows the sharing resources maintained by different administrative domains. The paper emphasizes the current stage of the platform and a case study in mobile robotics localization. Localization, as many other mobile robotics algorithms, can employ parallelism at the cluster and cloud levels in order to improve speed, reliability, and scaling.