Mobility-Aware Cluster Federated Learning in Hierarchical Wireless Networks

Implementing federated learning (FL) algorithms in wireless networks has garnered a wide range of attention. However, few works have considered the impact user mobility on performance. To fill this research gap, we develop theoretical model to characterize hierarchical (HFL) algorithm where mobile users may roam across edge access points (APs), leading incompletion inconsistent FL training. We provide convergence analysis conventional HFL with mobility. Our proves that performance deteriorates drastically highly-mobile users. And such decline will be exacerbated small number participants and large data distribution divergences among users’ local data. circumvent these issues, propose mobility-aware cluster (MACFL) by redesigning mechanism, update rule, aggregation scheme. also conduct experiments evaluate HFL, (CFL) simple averaging, our proposed MACFL. The results show MACFL can enhance performance, especially for three different cases: ( $i$ ) case non-independent identically distributed (non-IID) data, notation="LaTeX">$ii$ high mobility, notation="LaTeX">$iii$