Publication: Experimental evaluation of robot operating systems in distributed teleoperated scenarios
Loading...
Identifiers
Publication date
2022-09
Defense date
2022-10-13
Authors
Advisors
Journal Title
Journal ISSN
Volume Title
Publisher
Impact
Export
Abstract
The Robot Operating System (ROS) has been developed as an open-source framework
consisting of multiple libraries and tools to facilitate the design and development
of robot applications and components. Despite the significant benefits introduced by its
modular design, ROS suffers several drawbacks, including meeting real-time latency and
throughput constraints in non ideal-networking conditions, and on scalability in multirobot
deployments. To address these issues, a new version of ROS, dubbed as ROS 2, has
been recently released. ROS 2 benefits from several innovations, including Data Distribution
Service (DDS) and to enhance real-time communication features.
While ROS 2 may provide better performances in most of today’s realistic robotic
deployment, a constrained battery capacity still represent a major drawback in robotic
applications, specially those characterized by heavy computing of multiple data sources
such as video-analytics and machine-learning. In this context, 5G mobile networks as
well as edge/cloud computing paradigm offer the possibility to offload energy-demanding
computational tasks, such as Machine-Learning (ML) and video processing, to nearby
infrastructure, relieving the computational burden of the robots and increasing their operational
time.
In this thesis, we investigate the behavior of ROS and ROS 2 robotic application focusing
on remote teleoperated scenarios, comparing their performances in centralized and
distributed deployments, i.e., when processing is executed on the robot, or supported by
edge/cloud platforms by means of wireless communication.
As a result, we provide an exhaustive analysis of the generated traffic load exploiting
a virtualized environment, taking advantage of the measurement capabilities that such a
controlled scenario provides. Clearly, a trade-off exists between computational power,
networking usage, and battery consumption. Therefore, we consider these aspects and
provide a numerical evaluation characterizing the benefits of distributed systems over
centralized ones, also exploiting the latest ROS 2 technological improvements.
Description
Keywords
Robot operating system, 5G, Edge computing, Cloud computing