Interlinked Chain Method for Blockchain-Based Collaborative Learning in Vehicular Networks

Collaborative edge learning has emerged as a fundamental paradigm shift in vehicular technology, improving the efficiency and privacy of machine learning in edge environments. Nevertheless, the prevailing blockchain-based methods for collaborative learning have faced challenges such as communication delays and network congestion, mainly due to the continuous exchange of local models. In response to these challenges, this paper proposed an innovative interlinked chain method (ICM). The ICM uses interconnected sub-networks in which selected vehicles serve as inner aggregators. This setup ensures that only global models are transmitted on the blockchain, optimizing communication infrastructures and minimizing overhead. Our experimental results show that the proposed ICM outperforms conventional approaches by exhibiting higher efficiency and requiring significantly less communication exchange compared to traditional clustered and unified architectures, and reducing the communication overhead by more than 70%. Remarkably, the interlinked architecture avoids the performance degradation of collaborative learning and maintains consistency even when different federated schemes are implemented.