Abstract
Rotor-efficiency loss and thrust saturation reduce the attitude moments a quadrotor can deliver, so fixed reaching gains may request unavailable torque. This study builds on fast terminal sliding-mode control and radial basis function neural-network compensation with a saturation-aware constant power reaching law. A consistently regularized sliding surface and neural adaptation are combined with fault-aware allocation and reaching gains scheduled by pre-clipping rotor-thrust margins. A weighted-energy analysis retains rotor clipping and establishes regional exponential stability for hover after the effectiveness estimate is updated to its true value, under the theorem’s assumptions. In attitude-control simulations of a damaged maneuver, the schedule matches the fixed-high peak error under exact-state feedback. With continuous correlated gyro errors, its mean paired integral attitude error is 18.42% below the high-gain result and 2.50% above the low-gain result. It also reduces actual angular-rate RMS during the final hold relative to high gains. These simulation results demonstrate a balance between fault response and noise sensitivity under the tested attitude-control conditions without retuning the gains.