Publications Journal article · Mega-constellations · 2026
End-to-End Testbed Validation of Periodic and Probabilistic-Aware Constellation Scheduling for Non-Terrestrial Networks
Elsevier Acta Astronautica, vol. 250, pp. 69–83, 2026
Read on UPCommonsfreePublisher’s versionfree
Abstract
Integrated satellite-and-network scheduling frameworks for Non-Terrestrial Networks (NTN) are validated almost exclusively through simulation, in which task execution is instantaneous and data volumes match planned values exactly. Real operations deviate: data generation rates fluctuate, onboard resources drift from their modeled states, and constraint violations cascade into lost service time. To our knowledge, no existing testbed closes the scheduling loop of an integrated satellite-and-network scheduler under execution deviations. This work makes two contributions to address this gap. First, we propose an evaluation paradigm for execution-deviation correction methods in this class of schedulers, covering both proactive and reactive strategies, and implement it as a closed-loop, in-the-loop testbed in which a ground scheduler issues plans, a satellite-side execution segment runs them under controlled data-generation deviations, and telemetry from the executed schedule feeds the next planning cycle. Second, we design and implement the satellite-side execution segment that the paradigm requires: a containerized execution agent on satellite-representative hardware that runs scheduled tasks, monitors onboard resources through filesystem telemetry, enforces storage constraints through a recovery mode, and reports actual resource states back to the ground. We instantiate the loop with the Constellation Management System (CMS) as the ground scheduler under test, applying two complementary uncertainty-handling strategies it embeds-periodic rescheduling with telemetry feedback (reactive) and probabilistic-aware planning through a certainty factor (proactive)-across sixteen experiments. The results show that the two strategies address different stages of the same failure mode: rescheduling corrects resource drift after it has accumulated, while the certainty factor removes from the plan the tasks that the probabilistic traffic model expects to cause drift. Each lever also carries an operational cost-additional ground contacts for rescheduling, foregone collection capacity for the certainty factor-so the operating point that balances them is mission-specific. The testbed exposes both effects and lets the same characterization be reproduced for any mission configuration to locate its own operating point.
Cite
DOI 10.1016/j.actaastro.2026.08.074
BibTeX
@article{singla2026e2etestbed,
author = {Singla, Arnau and Calveras, Anna and Fraire, Juan A. and Ruiz-de-Azua, Joan A.},
title = {End-to-End Testbed Validation of Periodic and Probabilistic-Aware Constellation Scheduling for Non-Terrestrial Networks},
journal = {Elsevier Acta Astronautica},
volume = {250},
pages = {69--83},
year = {2026},
doi = {10.1016/j.actaastro.2026.08.074}
}