Lyapunov-gated gain regulation for explainable adaptive flight control

Adaptive flight control improves tracking under disturbance. Unconstrained adaptation inflates gains and stresses actuators. A Lyapunov-gated gain regulation framework is presented for nonlinear pitch control. Gain updates are frozen when a Lyapunov-like error-energy trend rises. A prediction–realisation correlation index then scales the update magnitude. A post-hoc SHAP layer attributes each gain update to its decision features. Adaptation is driven by a composite cost combining tracking error, actuator total variation and control magnitude. Evaluation uses a rigid-body six-degree-of-freedom aircraft model restricted to the pitch channel under a sustained mid-flight load. Seven controllers are compared, including model-reference and L1 adaptive control at two filter bandwidths. The gate alone leaves cumulative gain inflation at 3.60. The gate combined with the correlation index reduces inflation to 0.016. Tracking RMSE is 0.067 at actuator total variation 83.3, on the L1 tracking–actuator frontier. Stability-gated regulation yields bounded, interpretable adaptation at parity with adaptive baselines.