HPC-enabled Blockchain-integrated Multimodal Federated Learning-based Autonomous Vehicles for Cyber-Physical Systems

In autonomous vehicle (AV) environments, cyber-physical systems (CPS) produce significant quantities of diverse data from connected cars, roadside units, sensors, and edge devices. Data confidentiality, communication overhead, scalability, and real-time decision-making are issues that classical centralized machine learning (ML) frameworks are required to deal with. To address these drawbacks, this research proposes a HPC-enabled Blockchain-integrated Multimodal Federated Learning-based Autonomous Vehicles for Cyber-Physical Systems (HBFLAVCPS), integrating edge computing, high-performance computing (HPC), blockchain-based trust management, and multimodal data aggregation. The framework employs federated meta-learning to support rapid personalization and adaptation to dynamic traffic conditions while preserving data privacy by keeping raw data at local nodes. Scheduled task orchestration within the HPC environment accelerates global model convergence and enhances computational efficiency. The integration of multimodal aggregation, edge intelligence, and decentralized trust mechanisms enables scalable, secure, and low-latency decision-making for AV ecosystems. Experimental analysis demonstrates improvements in model personalization, communication efficiency, privacy preservation, security, and overall system responsiveness compared with conventional centralized and standard federated learning approaches. The proposed HBFLAVCPS framework significantly enhances resistance to malicious updates and provides a robust foundation for next-generation intelligent AV-CPS applications requiring real time, privacy-preserving, and trustworthy collaborative intelligence. The result shows that better global model performance with an accuracy of 93% while reducing communication overhead by 27% and improving personalization accuracy by 14% when compared to standard federated learning strategies.