RCLF-QP: Quadrotor controller fends off unknown cyberattacks
A novel adaptive control law keeps drones stable even under unbounded malicious attacks.
Ensuring quadrotor safety under partial actuator failures, external disturbances, and malicious cyberattacks is a major challenge due to their underactuated and nonlinear dynamics. Building on prior work for complete loss of two opposing rotors, researchers from multiple US universities address the additional threat of unknown and unbounded cyberattacks. Their baseline proportional-derivative (PD) feedback with observer-based decoupling handles mismatched disturbances but remains vulnerable to attacks on pseudo-control channels.
To close this gap, they propose a Resilient Control Lyapunov Function-based Quadratic Program (RCLF-QP) that incorporates a real-time online adaptive compensational term into the conventional CLF framework. This QP-based constrained optimization allows new control objectives and constraints to be integrated without altering stability guarantees. Combined with a model-based extended state observer, the system mitigates both aerodynamic and wind-induced disturbances and maliciously injected attacks. Simulations in high-fidelity environments demonstrate that the RCLF-QP prevents trajectory divergence and system instability where the baseline controller fails, offering a systematic and extensible architecture for resilient quadrotor autonomy.
- Uses a quadratic program-based optimization to adaptively compensate for unknown and unbounded cyberattacks in real time.
- Extends prior fault-tolerant control for two-rotor loss to additionally handle maliciously injected attacks on pseudo-control channels.
- High-fidelity simulations confirm stability and trajectory maintenance under attacks that crash baseline PD feedback control.
Why It Matters
Enables secure drone operations in adversarial environments, critical for delivery, surveillance, and autonomous flight safety.