FogROS2-Config: A Toolkit for Choosing Server Configurations for Cloud Robotics

Cloud service providers provide over 50,000 distinct and dynamically changing set of cloud server options. To help roboticists make cost-effective decisions, we present FogROS2- Config, an open toolkit that takes ROS2 nodes as input and automatically runs relevant benchmarks to quickly return a menu of cloud compute services that tradeoff latency and cost. Because it is infeasible to try every hardware configuration, FogROS2-Config quickly samples tests a small set of edge- case servers. We evaluate FogROS2-Config on three robotics application tasks: visual SLAM, grasp planning. and motion planning. FogROS2-Config can reduce the cost by up to 20x. By comparing with a Pareto frontier for cost and latency by running the application task on feasible server configurations, we evaluate cost and latency models and confirm that FogROS2- Config selects efficient hardware configurations to balance cost and latency. Videos and code are available on the website https://sites.google.com/view/fogros2-config I. I N T RO D U C T I O N Many new robotics applications require powerful computa- tional resources (such as GPUs, FPGAs, and TPUs), making it impractical and uneconomical to deploy on onboard robot hardware. Recently, the emergence of cloud robotics allows robots to operate with more affordable onboard compute hardware by leveraging a cloud-based Software as a Service (SaaS) model of computing. Robots can access and pay for compute resources as needed, to reduce the cost of deployment and operation. For example, an average lifespan of a home vacuum robot is 6 years and it runs 20 minutes per week. If we run the same task on a much faster AWS cloud machine (t2.medium), it takes only $6.00 for 6 years. In contrast, a typical single-board computer (such as Raspberry Pi), with hardware and energy cost more than $100.00. While there are clear cost advantages in using the cloud, the performance-cost trade-off is understudied in cloud robotics due to the following four challenges: (1) Cloud service providers offer many different interfaces and pricing structures, presupposing that users already understand the provider-specific machine type required for their needs, (2) Providers typically employ coarse-grained monthly or hourly rates but update prices hourly. (3) The trade-off of application latency and operating cost is often unknown or variable: slower machine execution might lead to prolonged machine usage and potentially higher costs, (4) Robotics experts often ∗Equal Contribution 1Department of Electrical Engineering and Computer Sciences AThe AUTOLab at UC Berkeley (automation.berkeley.edu). 2Department of Industrial Engineering and Operations Research 1,2University of California, Berkeley, CA, USA 3Robotics Institute, Carnegie Mellon University Fig. 1: A Sample Use Case of FogROS2-Config. With FogROS2- Config, users only need to input application-level requirements, such as latency and per-request cost. FogROS2-Config automates the cloud machine selection by modeling the latency cost tradeoff and facilities the cost-effective cloud robotics machine selection. FogROS2-Config automatically provisions the cloud machines and enables unmodified ROS2 applications to run as if all components are on the local robot. know the high-level application needs but may lack specific domain knowledge about the optimal compute specifications and cloud machine types for those requirements. We present FogROS2-Config, an openly available cloud robotics platform that automatically models cost and per- formance trade-offs for unmodified ROS2 applications at a per-request granularity. FogROS2-Config automates the latency analysis using Skypilot [1], an inter-cloud broker that orchestrates heterogeneous providers of cloud services. With over 50,000 available selections of cloud instances, it is infeasible to model the cost and latency for each instance to compute the corresponding optimum. However, FogROS2- Config models cost and latency by running benchmark tests on a small set of edge-case servers. FogROS2-Config models the cost and latency trade-off of how application performance corresponds to different servers and how that is mapped to specific machine types with various cloud service providers. Directly based on the application-level requirements, FogROS2-Config helps roboticists choose the cloud service provider and hardware specification. 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 12083