C2FL: Clustered Continual Federated Learning under Spatial and Temporal Drift

2026-06-16

AI Summary

This work addresses the failure of distributed learning in collective adaptive systems caused by the confluence of data privacy constraints, spatial heterogeneity, and temporal drift. To tackle this challenge, the authors propose a decentralized continual federated learning framework that enables nodes to self-organize into learning groups through spatial clustering. The approach uniquely integrates spatial structure and temporal dynamics into a unified model, introducing a dwell-time-aware adaptive model averaging mechanism coupled with experience replay to mitigate catastrophic forgetting. Experimental results demonstrate that the proposed framework significantly outperforms standard federated learning methods under simulated spatiotemporal drift scenarios, effectively restoring the system’s collective adaptability in dynamic environments while preserving data privacy.

Abstract

Collective Adaptive Systems (CAS) increasingly rely on machine learning to let each node learn from locally sensed data, aligning its behavior with the surrounding environment. Scaling this intelligence, however, raises fundamental challenges: sensed data is often privacy-sensitive, preventing centralized collection; nodes are mobile, traversing regions where nearby nodes perceive similar phenomena while distant ones observe radically different conditions, creating natural spatial clusters; and these distributions evolve over time due to mobility, introducing temporal drift that makes local models progressively stale. These dynamics arise across domains - vehicular sensing, drone-based monitoring, smartphone crowdsensing - yet the interplay of privacy, spatial heterogeneity, and temporal drift severely undermines conventional learning strategies. Therefore, we propose C2FL, a fully distributed Federated Learning (FL) approach where nodes self-organize into learning groups through spatial clustering, reflecting the geographic structure of the environment. To counteract temporal drift, each node combines experience replay with a dwell-time-aware adaptive averaging step, progressively incorporating the regional consensus as it remains longer within the same area, while preserving previously acquired knowledge under evolving distributions. We evaluate our approach on synthetic experiments that systematically reproduce spatial and temporal shifts, showing that standard federated strategies degrade significantly under these conditions and that our method restores robust collective adaptation.

Problem

• privacy
• spatial heterogeneity
• temporal drift
• federated learning
• collective adaptive systems

Innovation

• Clustered Federated Learning
• Spatial Clustering
• Temporal Drift
• Experience Replay
• Dwell-time-aware Averaging