Toward LEO Satellite Network Systems for Instantaneous Detection of Environmental Changes

A team from the University of Alberta has demonstrated that Low Earth Orbit (LEO) satellite constellations, when equipped with onboard computing and deep learning computer vision pipelines, can achieve sub-minute information freshness for real-time environmental change detection. Their simulation framework modeled orbital dynamics, distributed processing, and network routing across 720 trials involving 12 real-world constellation configurations, including Starlink, Kuiper, Telesat, and OneWeb. Using Age of Information (AoI) as the key metric—measuring the time between data generation at a sensor and its delivery to a user—the system tested whether in-orbit edge computing and high-speed inter-satellite links (ISLs) could support near-instantaneous wildfire detection.

The results show that constellation design dramatically impacts AoI performance, with average values ranging from 66.5 seconds to over 6,300 seconds. The best-performing configurations achieved an average AoI below 70 seconds and a peak AoI under 100 seconds, meeting the timeliness required for real-time environmental monitoring. This work, published on arXiv, suggests that upcoming orbital data centers and edge computing can revolutionize disaster response by enabling satellite networks to process imagery directly in space rather than relying on ground relays, slashing detection delays from minutes to seconds.