arxiv: 2604.19958 · v1 · submitted 2026-04-21 · 💻 cs.DC · cs.OS· cs.SY· eess.SY

Recognition: unknown

Equinox: Decentralized Scheduling for Hardware-Aware Orbital Intelligence

Ansel Kaplan Erol, Divya Mahajan

Pith reviewed 2026-05-10 00:56 UTC · model grok-4.3

classification 💻 cs.DC cs.OScs.SYeess.SY

keywords decentralized schedulingsatellite constellationedge computingmarginal costbarrier functionload sheddingorbital systemsenergy-aware scheduling

0 comments

The pith

Equinox schedules satellite tasks by comparing each task's value to a local marginal cost that rises with battery drain, heat, or queue pressure.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces Equinox as a lightweight decentralized scheduler for Earth-observation satellites facing intermittent energy and the risk that early task execution will deplete batteries needed later. It compresses battery, thermal, and queue constraints into one state-dependent marginal cost derived from a barrier function that rises sharply near limits, so each satellite runs a task only when its value exceeds the current cost. This local signal also triggers offloading to neighbors whose cost is lower, producing value-ordered load shedding and distributed balancing without policies, routing, or global state. A multi-day simulation of a 143-satellite constellation, grounded in Jetson Orin Nano measurements, reports 20% higher scientific goodput and 31% higher image-processing throughput than priority scheduling, plus 2.2x mean battery reserves and 5.2x execution rate under high demand by shedding work rather than failing. Readers care because static or priority methods collapse under contention while this approach adapts locally.

Core claim

Equinox enables adaptive scheduling by compressing time-varying constraints, including battery charge, thermal headroom, and queue backlog, into a single state-dependent marginal cost of execution derived from a barrier function that rises sharply near safety limits. This local signal serves as a constellation-wide coordination primitive. Tasks execute only when their value exceeds the current cost, enabling value-ordered load shedding without explicit policies. If local costs exceed a neighbor's, tasks are dynamically offloaded over inter-satellite links, achieving distributed load balancing without routing protocols or global state.

What carries the argument

State-dependent marginal cost computed locally from a barrier function that rises sharply near safety limits on battery, thermal headroom, and queue backlog; it encodes instantaneous pressure and future risk to drive value-ordered execution and neighbor offloading.

Load-bearing premise

A single state-dependent marginal cost derived from a barrier function can accurately encode both instantaneous pressure and future risk across battery, thermal, and queue constraints, enabling correct value-ordered decisions and offloading without explicit policies or global state.

What would settle it

Re-running the 143-satellite multi-day simulation with task values randomized or set independently of scientific utility and checking whether the reported 20% goodput and 31% throughput gains disappear while battery-reserve benefits remain.

Figures

Figures reproduced from arXiv: 2604.19958 by Ansel Kaplan Erol, Divya Mahajan.

**Figure 1.** Figure 1: Scheduling for OEC satellites must maintain processing power for the 30% of orbit spent in eclipse. by ignoring future resource constraints, they risk depleting battery reserves needed for higher-value inference tasks that arrive later, resulting in suboptimal performance over the full orbit. Resource-based systems [19, 20] enforce battery or thermal limits and may redistribute load via inter-satellite lin… view at source ↗

**Figure 2.** Figure 2: Context of Eqinox to Earth-observation alternatives. Only Eqinox incorporates both hardware state and image priors to schedule Earth Observation workloads. 10 minutes per orbit), shared downlink bandwidth, and the sheer volume of imagery generated by modern constellations [8, 31]. As a result, time-critical events such as wildfires, floods, and maritime incidents cannot be processed in time when relying … view at source ↗

**Figure 3.** Figure 3: Overview of the Eqinox framework, illustrating per-satellite marginal execution cost, per-satellite local scheduling, and ISL arbitrage components within a decentralized runtime for distributed resource management in orbital edge systems. Battery factor. Battery state governs the system’s ability to survive future execution, particularly across eclipse. Unlike other resources, energy is not continuously re… view at source ↗

**Figure 4.** Figure 4: summarizes the central trade-off in our evaluation. The baselines occupy opposite ends of the return-resilience spectrum: Phoenix preserves battery almost perfectly but delivers relatively low scientific goodput, while Priority increases scientific return at the cost of chronic depletion. Eqinox moves this frontier outward. It is the only system that simultaneously achieves high scientific goodput and st… view at source ↗

**Figure 5.** Figure 5: Constellation-level resource behavior under sustained orbital load. (a) Eqinox maintains a stable battery regime with lower thermal stress than the baselines. (b) As battery declines, lower-value tasks are shed first, preserving reserve for higher-value work deeper into eclipse. instead trades a small reduction in per-image precision for much longer active operation, processing 31% more images overall tha… view at source ↗

**Figure 7.** Figure 7: Execution rate vs. battery reserve time under increasing task contention (2–16 tasks/image), with dashed trend lines. Eqinox sustains a strictly higher frontier at equal contention. execution becomes much more expensive (1.90× and 3.53×). Neighbor costs stay nearly constant across all tiers, showing that offloaded work is naturally absorbed by healthy, sunlit satellites rather than redistributed among str… view at source ↗

**Figure 6.** Figure 6: CDF of per-satellite inactivity, measured as the fraction of 10-minute windows in which no images are processed. Dashed lines mark fleet means. Eqinox substantially reduces dark observation windows, indicating more consistent coverage across the constellation. battery level (>0.86), it demonstrates greater variance in battery distribution, reflecting uneven load accumulation across sunlit satellites. Eqi… view at source ↗

**Figure 9.** Figure 9: Control signal ablation: scientific goodput vs. battery reserve time for each pricing variant (72 h fMoW). The original, multiplicative Eqinox achieves the best combined outcome; additive composition and battery-only signals collapse battery reserve without improving goodput. retains enough reserve to exploit the freed capacity. Network capacity amplifies value awareness but does not replace it. With fre… view at source ↗

**Figure 10.** Figure 10: Board power (W) during a 600-second live trace on Jetson Orin Nano Super. Red shading marks GPU-active inference bursts. Idle runtime power sits at 3.7 W. without value differentiation the system cannot shed lowpriority work early to preserve power. The full Eqinox design is the only variant that simultaneously achieves strong scientific return and strong operational reserve. 4.8 Hardware Validation [P… view at source ↗

**Figure 12.** Figure 12: Robustness to ISL link failures. Scientific Goodput remains stable through 50% link failure. Significant degradation (−34% Scientific Goodput) occurs only at complete ISL loss. 10 4 10 3 10 2 Barrier coefficient 14.0 14.5 15.0 Sci. Goodput (M SV/hr) 20 25 30 35 40 Mean Battery Level (%) [PITH_FULL_IMAGE:figures/full_fig_p019_12.png] view at source ↗

**Figure 11.** Figure 11: Sensitivity to barrier coefficient 𝛽 (24h, 4 tasks). Scientific Goodput varies by ±10% across a 100× range. Higher 𝛽 safely raises mean battery level (right axis) due to a self-correcting cost feedback loop. confirming that per-satellite load imbalance in baseline systems arises from geographic eclipse asymmetry, which free ISL eliminates. In the operational setting where ISL bandwidth carries real cost… view at source ↗

read the original abstract

Earth-observation satellites are emerging as distributed edge platforms for time-critical tasks, yet orbital scheduling remains challenged by intermittent energy harvesting and temporal coupling where eager execution risks future battery depletion. Existing schedulers rely on static priorities and lack mechanisms to adaptively shed work. We present Equinox, a lightweight, decentralized runtime for resource-constrained orbital systems. Equinox enables adaptive scheduling by compressing time-varying constraints, including battery charge, thermal headroom, and queue backlog, into a single state-dependent marginal cost of execution. Derived from a barrier function that rises sharply near safety limits, this cost encodes both instantaneous pressure and future risk. This local signal serves as a constellation-wide coordination primitive. Tasks execute only when their value exceeds the current cost, enabling value-ordered load shedding without explicit policies. If local costs exceed a neighbor's, tasks are dynamically offloaded over inter-satellite links, achieving distributed load balancing without routing protocols or global state. We evaluate Equinox using a multi-day simulation of a 143-satellite constellation grounded in physical Jetson Orin Nano measurements. Equinox improves scientific goodput by 20% and image-processing throughput by 31% over priority-based scheduling while maintaining 2.2x higher mean battery reserves. Under high demand, Equinox achieves 5.2x the execution rate of static scheduling by gracefully shedding work rather than collapsing under contention.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

Equinox gives a workable decentralized scheduler for orbital edge nodes by folding constraints into a barrier-derived marginal cost, but the evaluation leaves the function's derivation and robustness under-specified.

read the letter

The core idea here is turning battery, thermal, and queue state into one local signal that both decides whether to run a task and whether to hand it off. That signal comes from a barrier function that grows steep near limits, so satellites can shed load without central coordination or explicit policies. The simulation, run over multiple days on a 143-satellite constellation and calibrated to Jetson Orin Nano power and thermal data, reports 20% higher scientific goodput, 31% better image throughput, and 2.2x battery headroom versus priority scheduling, plus 5.2x execution rate under heavy load. Those numbers are the strongest part of the work; they come from hardware-grounded traces rather than pure theory, which makes the claims easier to take seriously than most orbital scheduling papers. What is actually new is the specific packaging of the barrier cost as a constellation-wide coordination primitive that also drives neighbor offloading over inter-satellite links. Prior decentralized schedulers exist in other domains, but the orbital constraints and the value-ordered shedding without routing protocols are a fresh combination. The soft spot is the barrier function itself. The abstract and evaluation description give no explicit equation, no parameter-selection procedure, and no sensitivity sweeps, so it is impossible to judge whether the reported gains depend on careful tuning or would survive different orbit geometries or task mixes. The central assumption—that one state-dependent cost can reliably encode both immediate pressure and future risk across all three constraints—remains plausible but untested in the provided material. This paper is aimed at researchers working on distributed edge systems for space or other highly constrained environments. It is worth sending to peer review because the hardware grounding and the concrete performance deltas give referees something solid to examine, even if the math needs more transparency and the experiments need ablation on the barrier parameters.

Referee Report

3 major / 2 minor

Summary. The paper presents Equinox, a decentralized scheduling runtime for Earth-observation satellites that compresses battery, thermal, and queue constraints into a single state-dependent marginal cost derived from a barrier function. Tasks execute only when their value exceeds this local cost, enabling value-ordered shedding; offloading to neighbors occurs when local costs are higher, achieving load balancing without global state or routing protocols. A multi-day simulation of a 143-satellite constellation, grounded in Jetson Orin Nano measurements, reports 20% higher scientific goodput, 31% higher image-processing throughput, and 2.2x mean battery reserves versus priority-based scheduling, plus 5.2x execution rate under high demand via graceful shedding.

Significance. If the central claims hold, Equinox demonstrates a lightweight, policy-free mechanism for adaptive, hardware-aware scheduling in energy-intermittent orbital edge systems. The barrier-function approach to encoding future risk in local decisions could generalize to other distributed resource-constrained platforms. The simulation's grounding in physical Jetson Orin Nano measurements is a concrete strength that ties results to real hardware behavior.

major comments (3)

[Abstract] Abstract: the marginal cost is asserted to be 'derived from a barrier function that rises sharply near safety limits,' yet no explicit equations, functional form, or parameter-selection procedure for the barrier function are supplied. This is load-bearing for the central claim that the single cost 'encodes both instantaneous pressure and future risk' and produces the reported gains independently of tuning.
[Evaluation] Evaluation section: quantitative claims (20% goodput, 31% throughput, 2.2x battery reserves, 5.2x execution rate) are presented from a multi-day simulation without error bars, number of runs, statistical tests, or sensitivity analysis on barrier-function parameters, leaving the performance improvements unverifiable and the robustness of the decentralized coordination unassessed.
[Method] Method description: the weakest assumption—that one state-dependent marginal cost suffices to capture battery, thermal, and queue constraints simultaneously for correct local decisions and offloading—is stated without formal justification, convergence arguments, or counter-example analysis, which is required to support the claim of 'distributed load balancing without routing protocols or global state.'

minor comments (2)

[Abstract] Abstract: the term 'value' in 'value exceeds the current cost' and 'value-ordered load shedding' is used without a definition or how task value is computed or assigned.
[Evaluation] Evaluation: clarify the exact mapping from Jetson Orin Nano power/thermal traces to the simulation model parameters.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback. The comments identify key areas where additional clarity, statistical support, and justification will strengthen the manuscript. We address each major comment below and will incorporate revisions to improve verifiability and rigor.

read point-by-point responses

Referee: [Abstract] Abstract: the marginal cost is asserted to be 'derived from a barrier function that rises sharply near safety limits,' yet no explicit equations, functional form, or parameter-selection procedure for the barrier function are supplied. This is load-bearing for the central claim that the single cost 'encodes both instantaneous pressure and future risk' and produces the reported gains independently of tuning.

Authors: We agree the abstract would benefit from greater specificity on this central mechanism. Section 3.1 of the manuscript derives the marginal cost from the standard logarithmic barrier B(s) = -∑_i log(1 - s_i / s_i^max) over the normalized state vector s (battery charge, thermal headroom, queue backlog). This form rises sharply near any safety limit and is strictly increasing in each dimension, directly encoding both current pressure and future risk. Parameters are selected from Jetson Orin Nano hardware measurements to maintain conservative safety margins. We will revise the abstract to reference this functional form concisely and expand the methods section with an explicit parameter-selection procedure. We will also add sensitivity results to the evaluation to demonstrate that gains hold across reasonable parameter ranges. revision: yes
Referee: [Evaluation] Evaluation section: quantitative claims (20% goodput, 31% throughput, 2.2x battery reserves, 5.2x execution rate) are presented from a multi-day simulation without error bars, number of runs, statistical tests, or sensitivity analysis on barrier-function parameters, leaving the performance improvements unverifiable and the robustness of the decentralized coordination unassessed.

Authors: This critique is fair and points to a genuine limitation in the current presentation. The reported figures come from a single multi-day trace. In the revised manuscript we will execute 20 independent runs using varied random seeds for task arrivals, orbital phasing, and demand patterns. We will report means accompanied by standard-error bars, apply paired statistical tests (Wilcoxon signed-rank) to confirm significance of the 20–31% gains and 2.2× battery improvement, and include a sensitivity sweep over barrier parameters (±20% on safety thresholds). These additions will make the quantitative claims verifiable and directly address robustness of the decentralized coordination. revision: yes
Referee: [Method] Method description: the weakest assumption—that one state-dependent marginal cost suffices to capture battery, thermal, and queue constraints simultaneously for correct local decisions and offloading—is stated without formal justification, convergence arguments, or counter-example analysis, which is required to support the claim of 'distributed load balancing without routing protocols or global state.'

Authors: We will add a dedicated justification subsection. Because the barrier is additive and strictly convex in each constraint, the resulting scalar cost is monotonically responsive to any increase in future risk; local execution and offloading decisions therefore remain correct without needing to reconstruct the full constraint vector. For load balancing we will sketch a potential-function argument showing that each offload strictly decreases the neighborhood maximum cost, driving the system toward an equilibrium in which no further beneficial offload exists. We will also discuss a counter-example involving high communication latency that could induce transient oscillation, while noting that our scheduled inter-satellite-link model precludes this case. These additions supply the requested formal grounding while remaining within the scope of a systems paper. revision: partial

Circularity Check

0 steps flagged

No significant circularity in derivation chain

full rationale

The paper introduces Equinox by defining a marginal cost from a barrier function to compress constraints into a local signal for scheduling and offloading decisions. Performance gains (20% goodput, 31% throughput, 2.2x battery) are reported as direct outcomes of multi-day constellation simulations grounded in Jetson Orin Nano measurements, compared against priority-based and static baselines. No equations, self-citations, uniqueness theorems, or fitted-parameter renamings appear in the provided text that would reduce the central claims to inputs by construction. The barrier function serves as an explicit design choice whose effects are validated externally via simulation rather than asserted as a mathematical necessity. The derivation remains self-contained.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

The central claim rests on the ability of a barrier function to compress battery, thermal, and backlog constraints into a usable local signal and on the assumption that value can be assigned to tasks independently of that signal.

free parameters (1)

barrier function parameters
The sharpness and scaling of the barrier near safety limits are not specified; these values determine the marginal cost and are likely chosen or fitted to produce the reported behavior.

axioms (2)

domain assumption Task value can be compared directly to the instantaneous marginal cost to decide execution.
Required for the value-ordered load-shedding rule stated in the abstract.
domain assumption Inter-satellite links are available and reliable enough for dynamic offloading when local cost exceeds a neighbor's cost.
Necessary for the distributed load-balancing claim.

pith-pipeline@v0.9.0 · 5551 in / 1709 out tokens · 68379 ms · 2026-05-10T00:56:27.750483+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

40 extracted references · 24 canonical work pages · 1 internal anchor

[1]

Acero, Jonathan Diaz, Ronald Hurtado-Velasco, Sergio Ramiro Gonzalez Bautista, Sonia Rincón, Francisco L

Ignacio F. Acero, Jonathan Diaz, Ronald Hurtado-Velasco, Sergio Ramiro Gonzalez Bautista, Sonia Rincón, Francisco L. Hernández, Ju- lian Rodriguez-Ferreira, and Jesus Gonzalez-Llorente. 2023. A Method for Validating CubeSat Satellite EPS Through Power Budget Analysis Aligned With Mission Requirements.IEEE Access11 (2023), 43316– 43332. doi:10.1109/ACCESS....

work page doi:10.1109/access.2023.3271596 2023
[2]

Blaise Agüera y Arcas, Travis Beals, Maria Biggs, Jessica V Bloom, Thomas Fischbacher, Konstantin Gromov, Urs Köster, Rishiraj Prava- han, and James Manyika. 2025. Towards a future space-based, highly scalable AI infrastructure system design. arXiv preprint arXiv:2511.19468.https://arxiv.org/abs/2511.19468

work page arXiv 2025
[3]

Han Cai, Chuang Gan, Tianzhe Wang, Zhekai Zhang, and Song Han
[4]

Once-for-all: Train one network and specialize it for efficient deployment,

Once-for-All: Train One Network and Specialize it for Efficient Deployment. arXiv:1908.09791 [cs.LG]

work page arXiv 1908
[5]

Luigi Capogrosso and Michele Magno. 2026. TinyML Enhances Cube- Sat Mission Capabilities. arXiv:2603.20174 [cs.CV]https://arxiv.org/ abs/2603.20174

work page arXiv 2026
[6]

Chaudhry and Halim Yanikomeroglu

Aizaz U. Chaudhry and Halim Yanikomeroglu. 2021. Laser In- tersatellite Links in a Starlink Constellation: A Classification and Analysis.IEEE Vehicular Technology Magazine16 (2021), 48–56. doi:10.1109/mvt.2021.3063706

work page doi:10.1109/mvt.2021.3063706 2021
[7]

Gordon Christie, Neil Fendley, James Wilson, and Ryan Mukherjee
[8]

InProceedings of the IEEE Confer- ence on Computer Vision and Pattern Recognition (CVPR)

Functional Map of the World. InProceedings of the IEEE Confer- ence on Computer Vision and Pattern Recognition (CVPR). 6172–6180. doi:10.1109/CVPR.2018.00646

work page doi:10.1109/cvpr.2018.00646 2018
[9]

Google Coral. 2019. USB Accelerator Datasheet.https://coral.ai/docs/ accelerator/datasheet/

2019
[10]

Bradley Denby, Krishna Chintalapudi, Ranveer Chandra, Brandon Lu- cia, and Shadi Noghabi. 2023. Kodan: Addressing the Computational Bottleneck in Space. InProceedings of the 28th ACM International Con- ference on Architectural Support for Programming Languages and Oper- ating Systems, Volume 3 (ASPLOS 2023). Association for Computing Ma- chinery, New York...

work page doi:10.1145/3582016.3582043 2023
[11]

Bradley Denby and Brandon Lucia. 2020. Orbital Edge Computing: Nanosatellite Constellations as a New Class of Computer System. InProceedings of the Twenty-Fifth International Conference on Archi- tectural Support for Programming Languages and Operating Systems (ASPLOS ’20). Association for Computing Machinery, New York, NY, USA, 939–954. doi:10.1145/33733...

work page doi:10.1145/3373376.3378473 2020
[12]

Ansel Kaplan Erol, Seungjun Lee, and Divya Mahajan. 2025. Earth- Sight: A Distributed Framework for Low-Latency Satellite Intelligence. arXiv preprint arXiv:2511.10834.https://arxiv.org/abs/2511.10834

work page internal anchor Pith review Pith/arXiv arXiv 2025
[13]

European Space Agency. 2025. Sentinel Satellite Data.https://sentinel. esa.int/web/sentinel/user-guidesAccessed: 2025

2025
[14]

Corrado Gini. 1921. Measurement of Inequality of Incomes.The Economic Journal31, 121 (1921), 124–125

1921
[15]

Yu, Stephen J

Jianping Gou, B. Yu, Stephen J. Maybank, and Dacheng Tao. 2020. Knowledge Distillation: A Survey.International Journal of Com- puter Vision129 (2020), 1789 – 1819.https://api.semanticscholar. org/CorpusID:219559263

2020
[16]

Yifei Hu, Wenbin Gong, and Fangming Zhou. 2023. A Lyapunov- Optimized Dynamic Task Offloading Strategy for Satellite Edge Com- puting.Applied Sciences13, 7 (2023), 4281. doi:10.3390/app13074281

work page doi:10.3390/app13074281 2023
[17]

Longbo Huang and Michael J. Neely. 2013. Utility Optimal Sched- uling in Energy-Harvesting Networks.IEEE/ACM Transactions on Networking21, 4 (2013), 1117–1130. doi:10.1109/TNET.2012.2226406

work page doi:10.1109/tnet.2012.2226406 2013
[18]

Rachel Jewett. 2024. Planet to Use Nvidia AI Proces- sor Onboard Pelican Satellite.Via Satellite(June 2024). https://www.satellitetoday.com/innovation/2024/06/10/planet- to-use-nvidia-ai-processor-onboard-pelican-satellite/Accessed: 2024-06-10

2024
[19]

Leela S Karumbunathan. 2022. Nvidia Jetson AGX Orin Series. NVIDIA Technical Brief.https://www.nvidia.com/content/dam/en- zz/Solutions/gtcf21/jetson-orin/nvidia-jetson-agx-orin-technical- brief.pdf

2022
[21]

Zhouyu Li, Zhijin Yang, Huayue Gu, Xiaojian Wang, Yuchen Liu, and Ruozhou Yu. 2025. OrbitChain: Orchestrating In-orbit Real-time Analytics of Earth Observation Data. arXiv preprint arXiv:2508.13374. https://doi.org/10.48550/arXiv.2508.13374

work page doi:10.48550/arxiv.2508.13374 2025
[22]

Weisen Liu, Zeqi Lai, Qian Wu, Hewu Li, Qi Zhang, Zonglun Li, Yuanjie Li, and Jun Liu. 2024. In-Orbit Processing or Not? Sunlight-Aware Task Scheduling for Energy-Efficient Space Edge Computing Networks. In IEEE INFOCOM 2024 — IEEE Conference on Computer Communica- tions. IEEE, Vancouver, BC, Canada. doi:10.1109/INFOCOM52122. 2024.10621268

work page doi:10.1109/infocom52122 2024
[23]

NASA Atmospheric Science Data Center. 2025. CER SSF Aqua-FM3- MODIS Edition4A.https://asdc.larc.nasa.gov/project/CERES/CER_ SSF_Aqua-FM3-MODIS_Edition4A

2025
[24]

Planet Labs Inc. 2025. Our Constellations: Dove, RapidEye, and SkySat Fleets.https://www.planet.com/our-constellations/Accessed: 2025- 04-13

2025
[25]

Thomas Pusztai, Cynthia Marcelino, and Stefan Nastic. 2024. Hyper- Drive: Scheduling Serverless Functions in the Edge-Cloud-Space 3D Continuum. In2024 IEEE/ACM Symposium on Edge Computing (SEC). IEEE, Rome, Italy. doi:10.1109/SEC62691.2024.00028

work page doi:10.1109/sec62691.2024.00028 2024
[26]

2019.Skyfield: High Precision Research-Grade Posi- tions for Planets and Earth Satellites Generator.https://rhodesmill.org/ skyfield/

Brandon Rhodes. 2019.Skyfield: High Precision Research-Grade Posi- tions for Planets and Earth Satellites Generator.https://rhodesmill.org/ skyfield/

2019
[27]

Jinhyun So, Kevin Hsieh, Behnaz Arzani, Shadi Noghabi, Salman Aves- timehr, and Ranveer Chandra. 2022. Fedspace: An Efficient Federated Learning Framework at Satellites and Ground Stations. arXiv preprint arXiv:2202.01267.https://arxiv.org/abs/2202.01267

work page arXiv 2022
[28]

Spire Global Inc. 2025. Canadian Space Agency Assigns Can$72 Million Contract to Spire Global Canada to Design WildFireSat Mission.https://ir.spire.com/news-events/press-releases/detail/242/ canadian-space-agency-assigns-can72-million-contract-toAccessed: 2025-04-13

2025
[29]

Spire Global Inc. 2025. Spirepedia: Spire’s Satellite Constellation. https://spire.com/spirepedia/constellation/Accessed: 2025-04-13

2025
[30]

Mingxing Tan and Quoc V. Le. 2020. EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks. arXiv:1905.11946 [cs.LG] https://arxiv.org/abs/1905.11946

work page arXiv 2020
[31]

Bill Tao, Om Chabra, Ishani Janveja, Indranil Gupta, and Deepak Va- sisht. 2024. Known Knowns and Unknowns: Near-realtime Earth Observation Via Query Bifurcation in Serval. In21st USENIX Sym- posium on Networked Systems Design and Implementation (NSDI 24). USENIX Association, Santa Clara, CA, USA, 809–824.https://www. 13 usenix.org/conference/nsdi24/prese...

2024
[32]

Vallado, Paul Crawford, Richard Hujsak, and T.S

David A. Vallado, Paul Crawford, Richard Hujsak, and T.S. Kelso
[33]

InAIAA/AAS Astrodynamics Specialist Conference and Exhibit

Revisiting Spacetrack Report #3. InAIAA/AAS Astrodynamics Specialist Conference and Exhibit. AIAA. doi:10.2514/6.2006-6753

work page doi:10.2514/6.2006-6753 2006
[34]

Deepak Vasisht, Jayanth Shenoy, and Ranveer Chandra. 2021. L2D2: Low Latency Distributed Downlink for LEO Satellites. InProceedings of the 2021 ACM SIGCOMM Conference. Association for Computing Ma- chinery, New York, NY, USA, 151–164. doi:10.1145/3452296.3472932

work page doi:10.1145/3452296.3472932 2021
[35]

2020 , pages =

Pauli Virtanen, Ralf Gommers, Travis E. Oliphant, Matt Haberland, Tyler Reddy, David Cournapeau, Evgeni Burovski, Pearu Peterson, Warren Weckesser, Jonathan Bright, et al. 2020. SciPy 1.0: Fundamental Algorithms for Scientific Computing in Python.Nature Methods17 (2020), 261–272. doi:10.1038/s41592-019-0686-2

work page doi:10.1038/s41592-019-0686-2 2020
[36]

Nair, and Divya Mahajan

Irene Wang, Prashant J. Nair, and Divya Mahajan. 2023. FLuID: mit- igating stragglers in federated learning using invariant dropout. In Proceedings of the 37th International Conference on Neural Informa- tion Processing Systems(New Orleans, LA, USA)(NIPS ’23). Curran Associates Inc., Red Hook, NY, USA, Article 3202, 16 pages

2023
[37]

Jinfeng Wen, Jianshu Zhao, Zixi Zhu, Xiaomin Zhang, Qi Liang, Ao Zhou, and Shangguang Wang. 2025. SateLight: A Satellite Application Update Framework for Satellite Computing. InProceedings of the 40th IEEE/ACM International Conference on Automated Software Engineering (ASE). IEEE, Seoul, Republic of Korea. doi:10.1109/ASE62289.2025. 11334499arXiv:2509.12809

work page doi:10.1109/ase62289.2025 2025
[38]

Ardagna, Shangguang Wang, and Mengwei Xu

Chen Yang, Qibo Sun, Qiyang Zhang, Hao Lu, Claudio A. Ardagna, Shangguang Wang, and Mengwei Xu. 2024. Toward Efficient Satellite Computing Through Adaptive Compression.IEEE Transactions on Services Computing17, 6 (2024), 4411–4422. doi:10.1109/TSC.2024. 3470341

work page doi:10.1109/tsc.2024 2024
[39]

Su Yao, Yiying Lin, Mu Wang, Ke Xu, Mingwei Xu, Changqiao Xu, and Hongke Zhang. 2025. LEOEdge: A Satellite-Ground Cooperation Platform for AI Inference in Large LEO Constellations.IEEE Journal on Selected Areas in Communications43, 1 (2025), 36–50. doi:10.1109/ JSAC.2024.3460083

work page arXiv 2025
[40]

Xinyuan Zhang, Jiang Liu, Ran Zhang, Yudong Huang, Jincheng Tong, Ning Xin, Liang Liu, and Zehui Xiong. 2024. Energy-Efficient Computation Peer Offloading in Satellite Edge Computing Networks. IEEE Transactions on Mobile Computing23, 4 (2024), 3077–3091. doi:10.1109/TMC.2023.3269801

work page doi:10.1109/tmc.2023.3269801 2024
[41]

Yihua Zhang, Yuguang Yao, Parikshit Ram, Pu Zhao, Tianlong Chen, Min-Fong Hong, Yanzhi Wang, and Sijia Liu. 2022. Advancing Model Pruning via Bi-level Optimization.ArXivabs/2210.04092 (2022).https: //api.semanticscholar.org/CorpusID:252780187 14 A Experimentation Environment With the mechanism and its parameters defined, this section documents the physica...

work page arXiv 2022