Edge Datacenters to bring Cloud-hosted foundation models to devices
Drones and other robots rely on computer vision to understand their environment and accomplish their tasks, e.g., tracking targets and interacting with human users. Such tasks are often latency-sensitive (and sometimes mission-critical) and compute-intensive, often beyond the limited computing capacity of the local systems.
We aim at providing ML services to such autonomous systems from edge datacenters. We tackle four important systems challenges: Efficiency, Latency, Privacy, and Resilience. We leverage an edge-powered drone testbed developed at Athena and drive the research investigations with application cases in personal robotic assistance and collaborative target tracking.
