Eric Chen

EECS Department, University of California, Berkeley

Technical Report No. UCB/EECS-2024-70

May 9, 2024

http://www2.eecs.berkeley.edu/Pubs/TechRpts/2024/EECS-2024-70.pdf

Connectivity between robots and cloud machines is one of the most fundamental components in Cloud Robotics. We study the network infrastructure of FogROS2, an open-source Cloud Robotics platform, and propose FogROS2-SGC, a network extension of FogROS2 that can effectively connect robot systems across different physical locations, networks, and Data Distribution Services (DDS). FogROS2-SGC uses globally unique and location-independent identifiers to identify robots and cloud services. It securely and efficiently connects them and routes data between robotics components around the globe with a peer-to-peer network (Global Data Plane). FogROS2-SGC is agnostic to the ROS2 distribution and configuration, is compatible with non-ROS2 software, and seamlessly extends existing ROS2 applications without any code modification. Since memory copy and synchronization operations are expensive for memory-constrained robots, the implementation of FogROS2-SGC processes can route data without performing unnecessary copies (also known as “zero copy”). We consider latency critical cloud robotics applications that varying network conditions can cause instability and collisions. We observe such failure can be minimized in the almost universal case where there are multiple sources available for cloud servers. We extend FogROS2-SGC routing infrastructure by enabling multiple cloud robotics services to connect as the same globally unique identifiers. In the presence of multiple identical services, FogROS2-SGC dynamically identifies and transitions (“anycast”) to the optimal service deployment that meets latency requirements, thus empowering robots with limited on-board computing capacity to safely and efficiently navigate dynamic, human-dense environments. The anycast is achieved through managing a state machine- based scheduler that dynamically monitors the application latency and adjusts the routing states. We evaluate FogROS2-SGC with various simulated Cloud Robotics benchmarks (visual SLAM, motion planning, grasp planning). We show FogROS2-SGC with one connectivity case study with 4 robots and compute nodes that are 3600 km apart, and two latency sensitive case studies on collision avoidance and target tracking.

Advisors: John D. Kubiatowicz

---

BibTeX citation:

@mastersthesis{Chen:EECS-2024-70,
    Author= {Chen, Eric},
    Title= {FogROS2-SGC: A Secure and Global Connectivity Framework For Latency Sensitive Cloud Robotics},
    School= {EECS Department, University of California, Berkeley},
    Year= {2024},
    Month= {May},
    Url= {http://www2.eecs.berkeley.edu/Pubs/TechRpts/2024/EECS-2024-70.html},
    Number= {UCB/EECS-2024-70},
    Abstract= {Connectivity between robots and cloud machines is one of the most fundamental components in Cloud Robotics. We study the network infrastructure of FogROS2, an open-source Cloud Robotics platform, and propose FogROS2-SGC, a network extension of FogROS2 that can effectively connect robot systems across different physical locations, networks, and Data Distribution Services (DDS). FogROS2-SGC uses globally unique and location-independent identifiers to identify robots and cloud services. It securely and efficiently connects them and routes data between robotics components around the globe with a peer-to-peer network (Global Data Plane). FogROS2-SGC is agnostic to the ROS2 distribution and configuration, is compatible with non-ROS2 software, and seamlessly extends existing ROS2 applications without any code modification. Since memory copy and synchronization operations are expensive for memory-constrained robots, the implementation of FogROS2-SGC processes can route data without performing unnecessary copies (also known as “zero copy”).
We consider latency critical cloud robotics applications that varying network conditions can cause instability and collisions. We observe such failure can be minimized in the almost universal case where there are multiple sources available for cloud servers. We extend FogROS2-SGC routing infrastructure by enabling multiple cloud robotics services to connect as the same globally unique identifiers. In the presence of multiple identical services, FogROS2-SGC dynamically identifies and transitions (“anycast”) to the optimal service deployment that meets latency requirements, thus empowering robots with limited on-board computing capacity to safely and efficiently navigate dynamic, human-dense environments. The anycast is achieved through managing a state machine- based scheduler that dynamically monitors the application latency and adjusts the routing states.
We evaluate FogROS2-SGC with various simulated Cloud Robotics benchmarks (visual SLAM, motion planning, grasp planning). We show FogROS2-SGC with one connectivity case study with 4 robots and compute nodes that are 3600 km apart, and two latency sensitive case studies on collision avoidance and target tracking.},
}

EndNote citation:

%0 Thesis
%A Chen, Eric 
%T FogROS2-SGC: A Secure and Global Connectivity Framework For Latency Sensitive Cloud Robotics
%I EECS Department, University of California, Berkeley
%D 2024
%8 May 9
%@ UCB/EECS-2024-70
%U http://www2.eecs.berkeley.edu/Pubs/TechRpts/2024/EECS-2024-70.html
%F Chen:EECS-2024-70