Control adaptativo de vuelo de UAV basado en flujo de información y modelado aproximado para generalización de perturbaciones

Unmanned aerial vehicle (UAV) flight control systems are exposed to diverse security threats, including environmental disturbances, component faults, and cyber-attacks. Existing fault-tolerant control methods lack generalizability in representing such heterogeneous perturbations, particularly adversarial cyber-attacks, due to their reliance on type-specific modeling assumptions. In this study, an Information Flow Metamodel-Based Adaptive Approximate Modeling (IFM-AAM) framework is proposed to address this issue. The framework establishes an Information Flow Metamodel (IFM) as a high-level structural abstraction for approximate modeling. IFM represents heterogeneous perturbations as symbolic anomalies within a unified information flow structure, enabling consistent propagation analysis and impact computation across diverse perturbation sources. The IFM design supports flexible instantiation of domain models (DMs) at appropriate abstraction levels, addressing the computational challenges of modeling densely coupled flight control systems. In addition, a lightweight Actor-Critic reinforcement learning (RL) mechanism is integrated into the instantiated DMs for real-time adaptive tuning of control parameters, further enhancing modeling fidelity under perturbations. Comparative simulations against PX4 Autopilot demonstrate that DMs constructed by IFM-AAM effectively approximate control performance across nominal and diverse perturbation scenarios. These results confirm the effectiveness of our unified representation and propagation analysis for heterogeneous perturbations. Moreover, validations on an onboard computer verify the real-time performance and low computational overhead of DMs.