An Edge Cloud IoT Framework with Explainable AI for Real-Time Aircraft Cabin Monitoring and Predictive Safety

Abstract

Digitalization and increasing passenger safety requirements are continuously contributing to the growing complexity of modern aircraft cabins. Therefore, intelligent, realtime monitoring systems that can react quickly and foresee safety risks are required. Existing cabin monitoring solutions typically rely on centralized processing and rules-based logic, which have the drawbacks of latency, limited scalability, and lack of transparency in decision-making. In addition, the adoption of artificial intelligence (AI) in aeronautical safety is challenged by the ”black-box” nature of several high-performance models, raising concerns regarding trust, certification, and operational accountability. The present research proposes an edge-cloud Internet of Things (IoT) platform integrated with explainable artificial intelligence (XAI) for real-time aircraft cabin monitoring and predictive safety assessment. The proposed architecture utilizes distributed edge intelligence for low-latency anomaly detection and immediate response. In addition, cloud analytics supports long-term learning, cross-flight knowledge sharing, and predictive risk assessment. A multimodal sensing framework is employed to capture environmental, operational, and passengerrelated parameters within the cabin. The proposed framework was evaluated using a realistic sensor dataset representing cabin environmental monitoring scenarios. A statistical analysis of the experimental results shows that the proposed hybrid edge–fog–cloud architecture achieves substantial enhancements across multiple key performance indicators. In particular, the proposed model reduces average system latency by approximately 18-25%, decreases packet loss by nearly 15-20% under high network congestion conditions, and lowers transmission delay by approximately 12-18% compared with edge-only processing approaches. The hybrid architecture demonstrates improved computational efficiency by reducing average processing time by approximately 10% while maintaining a higher anomaly detection accuracy of up to 97–98%. Scalability analysis further indicates that the proposed framework maintains scalability efficiency above 0.98 even when the number of sensing nodes increases to 100, demonstrating strong robustness under largescale deployments. The integration of edge-cloud intelligence with explainable AI enhances situational awareness, operational transparency, and proactive decision-making in safety-critical aviation environments.