Adaptive Integral Sliding-Mode Control for UAV Hovering with Conditional Anti-Saturation

This paper develops conditional anti-saturation integral sliding-mode control for quadrotor hovering subject to bounded generalized control authority, external disturbances, and model uncertainties. A state-dependent Boolean mechanism conditionally modifies the bounded restoring action according to whether each hovering error is increasing or recovering. For the nominal hovering model, the resulting CAS-ISMC law establishes the integral sliding manifold from the initial instant under ideal initialization and guarantees asymptotic convergence of the hovering errors within a certified positively invariant recovery region, while an explicit gain budget enforces the prescribed componentwise input bounds. For the uncertain model, a projected adaptive extension estimates a structured uncertainty bound with unknown coefficients and known conservative limits. Projection prevents gain drift and yields a computable control-authority condition. Under a verifiable local initialization condition, the adaptive closed-loop signals remain bounded, the sliding variable converges to zero, and the hovering errors converge asymptotically to the origin. The discontinuous closed-loop dynamics is analyzed in the Filippov sense using Lyapunov arguments and weak-invariance analysis for the limiting differential inclusion. Reproducible simulations evaluate damping-gain sensitivity, projected compensation of structured matched uncertainty, and coupled four-channel near-hover regulation.