FogROS2-FT Fault Tolerant Cloud Robotics

Source

@inproceedings{ichnowski_2023_fogros2,
    author={Chen, Kaiyuan and Hari, Kush and Chung, Trinity and Wang, Michael and Tian, Nan and Juette, Christian and Ichnowski, Jeffrey and Ren, Liu and Kubiatowicz, John and Stoica, Ion and Goldberg, Ken},
    booktitle={IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)}, 
    title={{FogROS2-FT}: Fault Tolerant Cloud Robotics}, 
    year={2024},
    pages={1390--1397},
    doi={10.1109/IROS58592.2024.10802613}
}

(UC Berkeley)

IEEE

arXiv

Flash Reading

Abstract: Cloud computing can be costly and unreliable. FogROS2-FT mitigates these issues by providing a multi-cloud extension that replicates stateless robotics services, routes requests to these replicas, and directs the first response back. FogROS2-FT can switch between cloud providers based on the cost, availability, and QoS (Quality of Service).
Introduction: Many cloud providers offer spot VMs/instances to reduce price, but they are not stable sometimes. This system handles resource unavailability, oversubscription, and stochastic latency. FogROS2-FT simultaneously dispathes requests to identical services on multiple cloud providers, and returns the first response.
Related Work: Spot VMs are not considered in FogROS2 due to its instability. Fault tolerance systems are normally based on failover.
System Assumptions and Features: The process is assumed to be that a robot sending sensor data and awaiting control instructions. The services need to be stateless.

Extensions

Spot VMS, or spot instances, are a type of cloud computing resource that allows users to bid on unused capacity at a lower cost than regular on-demand instances. However, they can be terminated by the cloud provider with little notice, making them less reliable for critical applications.