A unified closed-loop stabilization framework for fixed-wing UAVs via natural-selection-enhanced multi-objective...

With the rapid expansion of civilian unmanned aerial vehicle applications, increasingly complex flight environments impose stricter requirements on flight stability, flying quality, and disturbance rejection. This study proposes a coupled lateral-longitudinal stabilisation framework for fixed-wing unmanned aerial vehicles based on a natural-selection multi-objective particle swarm optimisation strategy. A six-degree-of-freedom nonlinear flight dynamic model is developed under the rigid-body assumption and linearised around a trimmed steady-flight condition to derive lateral and longitudinal state-space models. According to flying quality theory, the Dutch roll mode in the lateral channel and the short-period and phugoid modes in the longitudinal channel are selected as performance indices. A unified closed-loop structure is constructed to address gain coupling between channels, and joint optimisation is performed using a natural-selection-enhanced multi-objective particle swarm optimisation algorithm. Actuator activity constraints and disturbance-response robustness criteria are incorporated to ensure control smoothness and engineering feasibility. MATLAB simulations show that the proposed method increases Dutch roll damping from 0.0766 to 0.4039 and improves the short-period damping ratio from 0.3868 to 0.8851 while satisfying Level-1 flying quality standards. Compared with the standard particle swarm optimisation algorithm, the proposed approach achieves faster convergence and higher constraint satisfaction. The optimised gains are directly applicable to the Pixhawk open-source flight control platform, providing practical guidance for low-cost stabilisation design in fixed-wing unmanned aerial vehicles. The novelty of the present work is fourfold: (i) lateral and longitudinal channels are tuned jointly rather than independently, so that gain-coupling between channels is explicitly resolved; (ii) a natural-selection mechanism is embedded in the MOPSO inner loop, which speeds up convergence by 22% versus standard MOPSO and 45-48% versus NSGA-II/III while raising the constraint-satisfaction rate to 98.5%; (iii) time-domain actuator-workload metrics (∫δ² and ∫(dδ/dt)²) are integrated alongside frequency-domain modal objectives, making the Pareto solutions directly deployable on bandwidth-limited Pixhawk-class servos; and (iv) the resulting framework is cross-validated on two distinct fixed-wing platforms (a conventional 13.5 kg Aerosonde-class UAV and a 1.56 kg Zagi-class flying wing) covering an order-of-magnitude range in mass and aspect ratio, confirming portability of the design procedure.