Inter-Locking Dependency Evaluation Schema based on Block-chain Enabled Federated Transfer Learning for Autonomous Vehicular Systems

Abstract

Federated Transfer Learning is a new approach to optimizing information and data training frameworks to achieve a higher order of distributed learning. Currently, the Internet of Vehicles (IoV) model is trained through online learning by exchanging the local and global updates. Context-based distributed learning with minimum communication becomes challenging due to the dynamic nature of autonomous systems. This paper aims to understand the past and future challenges, trends, and applications of new learning adopted in autonomous systems with a proposed schema of FTL-enabled autonomous models training. The goal behind minimizing the latency through the proposed FTL Interlocking Dependency Evaluation schema in exchanging the local and global updates is to match with the real-time application scenario by ensuring security and privacy features. The evaluation parameters used in estimating the performance of new learning are window size, back-off time, recovery policy (retransmission), scalability, peak data rate (Gbps), user-experienced data rate (Mbps), mobility (Kmph), latency (ms), end-to-end delay analysis, connection density (devices/km2), and area traffic capacity (Mb/s/m2). The evaluation results have validated the reliability of the proposed model for computing autonomous systems using a novel FTL Interlocking Dependency Evaluation schema for higher order of inter-dependency on vehicular environment processing.