The next generation of Low Earth Orbit (LEO) satellites, dedicated to Earth observation, seeks to enable a compelling vision: frequent high-resolution monitoring of the Earth to track humanity scale events. These satellites operate in low orbits, less than 1000 Km above Earth, and democratize access to Earth imagery for multiple applications: precision agriculture that improves farm yields and incomes, disaster management such as early detection of forest fires, modeling the spread of diseases, geo-political analytics, and climate monitoring. As such constellations scale to hundreds of satellites, traditional architectures for space-Earth connectivity fail to meet their needs. Specifically, these constellations generate hundreds of terabytes of data every day that must be transported to the Earth, while the satellites travel at fast speeds with respect to ground stations on the Earth. Therefore, the data download process experiences day-level delays and is prone to failures due to bad weather or hardware errors. The goal of this proposal is to design a new paradigm for space-Earth connectivity that removes these bottlenecks and enables fault-tolerant networks that generate near-realtime (minute-level) insights from data collected by these satellites. The proposed research will remove networking and compute bottlenecks for LEO satellites and enables novel commercial, academic, and national security applications. Some example applications include early response to aircraft intrusions, forest fires, and geopolitical events. The educational efforts in this proposal will train students to engage in research and development efforts focused on space-Earth networking and edge computing.
The proposed research designs new algorithms and architectures for networking and edge computing on satellites and ground stations. Specifically, the proposed research includes: (a) a new distributed ground station architecture that is robust to transient failures, agile to traffic variations, and enables low latency data download from satellites, (b) an edge computing framework that extracts insights, schedules data transfer, and prioritizes important imagery for networked transfer while being scalable to multiple applications, and (c) design and analysis of interconnects between communication and Earth observation satellites for opportunistic routing. The proposal takes an end-to-end systems level approach to large-scale space-Earth networks, which is essential for their robust, fault-tolerant, and high-performance operation.
Publications
- Known Knowns and Unknowns: Near-realtime Earth Observation Via Query Bifurcation in Serval. Bill Tao, Om Chabra, Ishani Janveja, Indranil Gupta, Deepak Vasisht. USENIX NSDI 2024.
- Code and dataset release: https://github.com/ConnectedSystemsLab/Umbra/
Students
Acknowledgments
This work is generously supported by the National Science Foundation through NSF CAREER grant, NSF CNS-2237474.