MLFCIL: A Multi-Level Forgetting Mitigation Framework for Federated Class-Incremental Learning in LEO Satellites

TL;DR

This paper introduces MLFCIL, a framework designed to mitigate catastrophic forgetting in federated class-incremental learning for LEO satellites, addressing unique challenges like data heterogeneity and resource constraints.

Contribution

MLFCIL decomposes forgetting into three sources and applies multi-level strategies to effectively mitigate it in resource-limited LEO satellite environments.

Findings

MLFCIL significantly improves accuracy over baselines.

MLFCIL effectively reduces forgetting in experiments.

The framework introduces minimal resource overhead.

Abstract

Low-Earth-orbit (LEO) satellite constellations are increasingly performing on-board computing. However, the continuous emergence of new classes under strict memory and communication constraints poses major challenges for collaborative training. Federated class-incremental learning (FCIL) enables distributed incremental learning without sharing raw data, but faces three LEO-specific challenges: non-independent and identically distributed data heterogeneity caused by orbital dynamics, amplified catastrophic forgetting during aggregation, and the need to balance stability and plasticity under limited resources. To tackle these challenges, we propose MLFCIL, a multi-level forgetting mitigation framework that decomposes catastrophic forgetting into three sources and addresses them at different levels: class-reweighted loss to reduce local bias, knowledge distillation with feature replay and prototype-guided drift compensation to preserve cross-task knowledge, and class-aware aggregation to mitigate forgetting during federation. In addition, we design a dual-granularity coordination strategy that combines round-level adaptive loss balancing with step-level gradient projection to further enhance the stability-plasticity trade-off. Experiments on the NWPU-RESISC45 dataset show that MLFCIL significantly outperforms baselines in both accuracy and forgetting mitigation, while introducing minimal resource overhead.