FogROS2-LS: A Location-Independent Fog Robotics Framework for Latency Sensitive ROS2 Applications

In Cloud Robotics, long system latency due to varying network conditions can cause instability and collisions. However, this can be minimized in the almost univeral case where there are multiple sources available for cloud servers. By extending anycast routing, we introduce FogROS2-Latency- Sensitive, a Fog Robotics framework that offers secure, location- independent connections between robots and latency-sensitive cloud-based servers. FogROS2-LS offloads conventional on- board state estimators and feedback controllers to Cloud and Edge compute hardware without modifying existing applica- tions in ROS2. In the presence of multiple identical services, FogROS2-LS dynamically identifies and transitions to the optimal service deployment that meets latency requirements, thereby empowering robots with limited on-board computing capacity to safely and efficiently navigate dynamic, human- dense environments. We evaluate FogROS2-LS with two latency sensitive case studies: (1) Collision Avoidance: a robot arm guided by visual feedback from consistent distance estimation and collision checking on Cloud and Edge. FogROS2-LS reduces collision failures by up to 8.5x by selecting the best available server, and (2) Target Tracking: FogROS2-LS enables robust and continuous target following and can recover from network failures. Videos and code are available on the website https://sites.google.com/view/fogros2-ls. I. I N T RO D U C T I O N Cloud, or Fog Robotics ([1], [2], [3]) enables robots to access external computing resources for (1) advanced visual perception ([4], [5], [6], [7], [8]); and (2) reinforcement learning-based intelligent motion control ([9], [10]). Our previous work introduced FogROS2—now an official part of the ROS2 ecosystem—which outsources heavy computing tasks to on-demand hardware resources and accelerators, such as GPU, TPU, ASIC, FPGA, and high performance CPU servers. A common misconception about cloud and fog robotics is that they are unsuitable for latency-sensitive and safety-critical tasks due to network failures and congestion. In this work, we assume the existence of multiple inde- pendent cloud compute servers and providers, and present FogROS2-Latency Sensitive, a Fog Robotics framework that enables reliable latency performances by dynamically selecting the optimal service out of all available servers. We evaluate its effectiveness in real robotics experiments with collision avoidance and continuous target tracking (Fig. 1). 1Department of Electrical Engineering and Computer Science 2Robert Bosch Research and Technology Center North America, Sunny- vale, CA, USA 3Department of Industrial Engineering and Operations Research 1,3University of California, Berkeley, CA, USA 4Robotics Institute, Carnegie Mellon University †For correspondence and questions: kych@berkeley.edu Fig. 1: A Sample Use Case of FogROS2-LS FogROS2-LS enables the location-independent deployment of fog robotics applications, allowing robots to connect with distributed robotic services with a unified ROS2 interface. It enables robust operation of latency sensitive applications, such as tracking or collision detection, by connecting robots to service deployments that satisfy these bounds. ROS2 [11] is the de-facto platform for building robotics applications in a location-independent way: heterogeneous robots and modular services1 publish to and subscribe from (pub/sub) each other as if they are running on the same machine. However, completely adhering to the ROS2 multiple- party pub/sub communication paradigm in fog robotics falls short on the following aspects: (1) Mirrored robotics services can be distributed to heterogeneous geographic locations and network domains. The framework needs to globally and securely discover and connect robots with those service deployments, while differentiating services hosted by other users or tasks. (2) The pub/sub paradigm leads to the request being published to all the deployments, leading to more network congestion and failure. While still adhering to ROS2 interfaces, a robot should select the one deployment that fulfills the application-specific latency bound. (3) Deployment selection should be adaptive to fluctuating application latency caused by varying network latency and hardware resource utilization, and network failures FogROS2-LS introduces a latency-aware location inde- pendent routing architecture. It enables launching multiple instances of robotics task servers across different geograph- ical and network domains, all identifiable by a location- independent identifier unique to the task. Robots use this 1Service in this paper refers to generic robotics application instead of the specific ROS2 service communication model. FogROS2-LS supports both publish-subscribe and service communication models in ROS2. 2024 IEEE International Conference on Robotics and Automation (ICRA 2024) May 13-17, 2024. Yokohama, Japan 979-8-3503-8457-4/24/$31.00 ©2024 IEEE 10581