AI-SAFE FlightNet : A Next-Generation Artificial Intelligence and Automation Framework for Predictive Flight Safety, Aircraft Health Intelligence, and Ground Operations Coordination

Research on Flight Anomaly Detection and Fault Diagnosis of Military Aircraft Based on Inertial Navigation

A Safety-Driven Architectural Framework for Fail-Operational Drone Swarms in Critical Missions

A mixed-criticality architectural framework that applies SAE ARP4754B methods to swarm reconfiguration, enabling intelligent swarm behaviors without compromising flight-critical isolation is proposed.

MulRobBench: A Decision-Level Benchmark for Safe and Security-Policy-Compliant Multimodal UAV Agents

MulRobBench provides a reproducible benchmark for trustworthy multimodal UAV decision making under realistic operational constraints and identifies modality-trust selection, constraint extraction, glare, missing data, and operator shorthand as the primary causes of decision instability.

Traceable LLM-Generated Hazard Scenarios for Operational Safety Analysis of Aviation Systems Using ASRS Reports

This work presents an AI-assisted approach that generates candidate hazard scenarios from NASA's Aviation Safety Reporting System (ASRS), and proposes a hybrid variant, conditioning narrative generation on a structured hypothesis produced via evolutionary abduction, improving correctness and reducing variability.

An integrated operational safety assurance method for reconfigurable flight control systems based on uncertainty quantification and resilience decision-making

The dynamic reconfigurability of civil aircraft flight control systems (FCS) enhances mission adaptability yet introduces significant operational uncertainty. Traditional static safety assurance methods often fail to cope with such real-time variations. To address this, this paper proposes an integrated safety assurance framework that synergistically combines uncertainty quantification, resilience engineering, and multi-criteria decision-making. First, this paper establishes a multidimensional uncertainty ontology encompassing environmental, system-internal, human–machine, and reconfiguration factors, which is quantified using hybrid probabilistic graphical models. Building on this foundation, this paper develops a closed-loop resilience architecture that embeds the four core capabilities of resilience engineering (anticipating, monitoring, responding, learning) to enable dynamic risk perception and adaptive reconfiguration. Specifically, this paper utilizes a time-evolving multi-attribute utility function coupled with online optimization to support worst-case-oriented robust decision-making. Central to the approach is a time-evolving multi-attribute utility function coupled with online optimization, which supports worst-case-oriented robust decision-making. Validation through a dual-disturbance scenario (severe weather and sensor degradation) demonstrates significant improvements in decision quality and operational resilience. Compared with a traditional fault tree analysis (FTA) baseline under identical simulation settings, the proposed framework demonstrates up to a 25%–35% improvement in the timeliness of risk state identification and an approximately 30–40% enhancement in decision robustness under sensor degradation and dynamic environmental uncertainty.

Cyber-resilient flight architecture for software-defined UAVs using digital twin-based validation

Software-Defined Unmanned Aerial Vehicles (SD-UAVs) rely heavily on software control and networked communication, making them vulnerable to cyber-physical attacks that threaten flight safety and mission reliability. Traditional UAV security approaches are largely reactive and lack integrated resilience mechanisms capable of sustaining stable flight under adversarial conditions. This study proposes a cyber-resilient flight architecture that integrates software-defined control, AI-driven anomaly detection, and Digital Twin-based validation. A Long Short-Term Memory (LSTM) model was implemented for telemetry anomaly detection, while SHAP-based Explainable AI was used to ensure interpretable resilience decisions. The system was implemented using Pixhawk 6X hardware, NVIDIA Jetson Orin Nano processing, MAVLink communication, and a Gazebo Garden + ROS2 Humble simulation environment. A Lyapunov-based stability analysis was conducted to validate the 50 m synchronization loop. Comparative evaluation demonstrated improved anomaly detection latency, reduced false detection rates, faster recovery time, and lower trajectory deviation compared to a conventional UAV architecture. The proposed system maintained stable flight under simulated cyberattack scenarios including command manipulation and sensor perturbation. The integration of Trustworthy AI with digital twin validation enhances UAV resilience, operational safety, and adaptive recovery under cyber threats, providing a scalable framework for secure autonomous flight systems.