ENHANCING TRUSTWORTHINESS OF IOT-ENABLED AUTOMATED VEHICLE LOCALIZATION SYSTEMS

Abstract

Relevance. Autonomous vehicles rely on multi-sensor localization systems operating within IoT infrastructures, creating interconnected vulnerabilities from sensor anomalies, network failures, and cybersecurity threats that require comprehensive solutions addressing both vehicle-level and infrastructure-level reliability challenges. The object of research is IoT-enabled automated vehicle localization systems requiring trustworthy operation under adverse conditions, including sensor malfunctions, GPS spoofing attacks, and infrastructure failures. Purpose of the article is to develop and validate a unified resilience framework that integrates transformer-based anomaly detection for in-vehicle sensor streams with federated learning agents deployed across IoT edge gateways, ensuring sub-second recovery from infrastructure failures while maintaining localization accuracy. Research results. The proposed framework achieves 9498% anomaly detection accuracy while maintaining localization errors below 0.5 meters during fault conditions. The federated learning component demonstrates 40% reduced communication overhead compared to centralized approaches, with sub-second failover capabilities during infrastructure failures. Explainable ML integration provides interpretable alerts through transformer attention mechanisms, enabling real-time system diagnostics. Conclusions. The unified framework successfully addresses critical challenges in autonomous vehicle deployment by combining multi-layer anomaly detection, coherent reliability broadcasting, and explainable AI techniques, providing a comprehensive foundation for trustworthy autonomous vehicle operation in IoT-enabled smart city environments.