Multirotor unmanned aerial vehicles (UAVs) suffer from significant control performance degradation during aggressive maneuvers, primarily due to aerodynamic nonlinearities and coupling effects. Conventional fixed-gain PID controllers struggle to simultaneously satisfy performance and robustness requirements across the wide flight envelope. To address this challenge, this paper presents a novel hierarchical safety-constrained reinforcement learning (RL) framework for adaptive PID tuning: the inner loop employs fixed gains, the outer loop leverages proximal policy optimization (PPO) for online adaptive gain scheduling, and linear matrix inequality (LMI) constraints delineate robust parameter boundaries for the adaptive exploration. Importantly, the LMI feasibility strictly guarantees theoretical stability for the fixed inner-loop parameters at the linearization vertices within a linear parameter-varying (LPV) framework. Concurrently, the online outer-loop RL stage is protected by safety boundaries and a Lagrangian penalty mechanism acting as an effective engineering safeguard rather than a rigorous global stability proof. Comprehensive high-fidelity simulation benchmarks demonstrate that, compared with a baseline fixed-gain PID controller, the proposed framework reduces overshoot by 18.5% in high-speed step responses and improves the overall mean RMSE by 15.0% across 100 randomized mixed-trajectory trials (with improvements of up to 40.9% in highly dynamic scenarios), yielding consistent gains in trajectory tracking accuracy and disturbance rejection despite uncertain model variations. By seamlessly blending control-theoretic hard constraints with RL-based soft-parameter tuning, the proposed architecture offers a safe and highly adaptive solution for large-envelope flight control, demonstrating strong engineering relevance.

Reliable local three-dimensional (3D) wind estimates are important for fixed-wing UAV flight under wind disturbances, but low-cost platforms lack direct 3D flow sensing. We propose PIRNN-AKF, which combines a physics-informed gated recurrent unit (PI-GRU) with a measurement-noise-adaptive Kalman smoother. PI-GRU embeds the wind triangle and attitude rotations in an airspeed-closure loss and predicts covariance scales for adaptive fusion. In PX4/JSBSim simulations with Dryden turbulence, five-seed PI-GRU attains a 3D RMSE of $0.208\pm 0.003$ m/s and a direction MAE of 3.22° on Test-ID. PIRNN-AKF attains $0.544\pm 0.007$ m/s and 1.97° on Test-OOD, reducing RMSE by 37.4% relative to Vanilla GRU; KalmanNet yields lower OOD RMSE but weaker ID accuracy and direction estimation. Session-aware evaluation shows that the AKF reduces OOD jitter by 24.8% while preserving RMSE. In ten frozen HITL sessions of the final 41-input, 100-step model, ID/OOD RMSEs are $0.569\pm 0.054$/$0.832\pm 0.318$ m/s, the companion-side processing $p_{95}$ is $20.4\pm 0.5$ ms, and the achieved loop rate is $83.2\pm 1.6$ Hz with $5.6\pm 0.7\%$ deadline misses. These results support 50 Hz feasibility under simulation truth; real-flight accuracy and closed-loop benefits remain untested.