Hierarchical Federated Transfer Learning for Digital Twin Vehicular Networks

A new paper on arXiv (2608.11532) proposes a Hierarchical Federated Transfer Learning (HFTL) framework to address data heterogeneity and sparsity in Digital Twin-based Vehicular Ad hoc Networks (DT-VANET). Federated Learning (FL) in such networks struggles to train a global model accurately due to varying data distributions across vehicle types. The authors combine Federated Transfer Learning (FTL) with vehicle clustering to improve model accuracy. They construct a DT-VANET framework and design two algorithms: one for cloud server model updates and another for intra-cluster federated transfer learning. Additionally, they introduce a data quality score-based mechanism to prevent malicious vehicles from degrading the global model. The paper details the framework and algorithms, aiming to enhance prediction accuracy for different vehicle types while preserving data privacy.