Robust Attitude Control of Fixed-Wing UAVs Under Near-Envelope Conditions Using a Hierarchical Adaptive NSGA-II

To address the strong aerodynamic coupling and insufficient control robustness of fixed-wing unmanned aerial vehicles (UAVs) under near-envelope conditions, this study focuses on three-axis attitude stabilization and develops a two-level control architecture consisting of an attitude/load-factor stability-augmentation loop and an angular-rate stability-augmentation loop. The outer loop adopts an attitude-angle dynamic-inversion and load-factor correction structure, whereas the inner loop employs linear active disturbance rejection control (LADRC). An improved non-dominated sorting genetic algorithm II (NSGA-II) integrating hierarchical pre-optimization and a threefold adaptive mechanism is proposed for multi-objective controller-parameter tuning. Hierarchical single-objective pre-optimization is followed by multi-objective coordinated optimization to reduce premature convergence in high-dimensional parameter optimization. A threefold adaptive mechanism comprising parameter-range adaptation, crossover and mutation probability adaptation, and objective-weight adaptation balances global exploration and local convergence. Nonlinear simulations and field tests show that hierarchical pre-optimization accounts for the main fitness reduction relative to standard NSGA-II (approximately 4.41%), whereas the adaptive mechanism lowers the post-convergence coefficient of variation from 0.0020 to 0.0006 and adds a 0.083% reduction beyond hierarchical pre-optimization alone. Under combined multi-source perturbations in aerodynamic, inertial, and actuator parameters, the optimized three-channel closed-loop controller remains stable within the tested ranges and maintains attitude-tracking performance. This study supports offline tuning under the tested near-envelope conditions.