| We explore the design of a nonlinear control system, enabling high performance maneuvers over the full flight envelope, tailored to embedded digital applications. The control law is designed in discrete-time using an sampled form of the continuous nonlinear dynamics of the aircraft, obtained from the Adams-Bashforth numerical method. Then, the resulting dynamics are decoupled into fast and slow dynamics and are controlled in the context of a successive loop closure using dynamic inversion. In order to provide robustness against perturbations due to model uncertainty, parameter variation, and external disturbances, a discrete-time sliding mode control (DSMC) is synthesized with dynamic inversion. The resulting DSMC does not require a priori knowledge of the perturbations and leads to a non-switching type of control. In order to account for hard limits on actuator range, a real-time control input limiting algorithm is also presented. This algorithm maps a given maximum control deflection boundary into an acceleration boundary in the state space via the aircraft dynamics, and modifies the tracking command vector such that the resulting acceleration vector for a given state always lies within the acceleration boundary. The resulting algorithm computes modified commands in polynomial time and guarantees the Lyapunov stability of the closed-loop system under saturation. Finally, we implement the proposed control scheme on one of our Stanford DragonFly UAVs and perform flight tests. The flight tests show that the control scheme successfully stabilizes and reduces tracking errors under rapid roll maneuvers, while keeping the control inputs within pre-defined bounds of saturation. |
