首页 > AI前沿 > Resource-Aware Federated Mixture-of-Experts with Adaptive Pruning for Onboard Learning in LEO Satellite Constellations

Resource-Aware Federated Mixture-of-Experts with Adaptive Pruning for Onboard Learning in LEO Satellite Constellations

arXiv机器学习 2026-09-26 03:24 5 阅读 查看原文

Low-Earth-orbit (LEO) satellites are increasingly expected to perform onboard learning for applications such as disaster response and environmental monitoring.

However, conventional federated learning (FL) is ill-suited to onboard satellite learning, as it assumes computational, memory, and communication resources beyond the capabilities of resource-constrained LEO platforms, often necessitating the transmission of raw imagery to ground stations.

COSMIC-FL: A Resource-Aware FL Framework

We present COSMIC-FL, a resource-aware FL framework for efficient onboard learning in LEO satellite constellations.

COSMIC-FL introduces two complementary Mixture-of-Experts (MoE) architectures:

  • A Sliced design that shares backbone representations while activating task-specific channel subsets,
  • A Modular design that employs lightweight gating to route inputs to physically separated expert networks.

A semantic class-to-expert mapping enables each satellite to train, update, and communicate only the expert paths relevant to its local data.

Efficiency Improvements

To further improve efficiency, COSMIC-FL integrates staged optimization with three structured pruning strategies:

  • Server-side pruning,
  • Client-side fixed-ratio pruning with mean-vote aggregation,
  • Adaptive client-side per-layer pruning based on aggregated importance and a MAD-based gap criterion.

Combined with semantic expert routing, these techniques jointly adapt computation and model sparsity to both data semantics and layer importance, yielding a favourable accuracy--efficiency trade-off for heterogeneous space platforms.

Experiments and Validation

Experiments on six image classification benchmarks under highly non-i.i.d. settings show that COSMIC-FL maintains competitive accuracy while reducing communication, computation, and energy consumption by up to 80% over SOTA FL methods.

We further validate COSMIC-FL on an NVIDIA Jetson AGX Orin, confirming its efficiency gains under realistic embedded deployment constraints.