The PoundCloud framework for ROS-based cloud robotics: Case studies on autonomous navigation and human–robot interaction

Abstract

Despite the increasing popularity of the cloud robotics paradigm, the literature on the field still lacks comprehensive analysis on several aspects of the technology. Therefore, the adoption of common standards and frameworks is fundamental for developing the field and allowing practical works to be reproduced and compared. This work presents a ROS-based open framework for robot-cloud communication, easing the integration of robotics and remote cloud platforms, and discusses the implementation of the overall cloud robotics stack over open source software and commercial off-the-shelf devices. Additionally, we present two practical implementations in which most of the computation is carried out remotely and perform a series of experiments to demonstrate our technique. Our results indicate that task times can be reduced up to 15% when using remote cloud platforms even under 150 ms average communication latency over the public Internet while observing figures as low as 2% on throughput loss in sensor data transmission. In general, such results point to the feasibility of the presented approach in different classes of applications, even under non-ideal network and cloud settings.