Robust $\mathcal{H}_\infty$ Controller Design For INDI-Controlled Quadrotor Using Online Parameter Identification

A team of researchers led by Tom Aantjes has developed a robust gain-scheduled controller for quadrotors that combines Incremental Nonlinear Dynamic Inversion (INDI) with H-infinity closed-loop shaping. The key innovation is the ability to identify all physical parameters of the quadrotor onboard in under half a second – fast enough to complete the identification during a throw in the air. Until now, a robust method to tune the outer loop gains for this feedback-linearizing INDI controller based on the identified model parameters was missing.

The proposed solution uses signal-based H-infinity synthesis to design a gain-scheduled cascaded attitude controller with a feedforward filter. Nonlinear simulations confirm that the resulting controller exhibits good stability margins and effective tracking performance under uncertainty. Experimental flight tests with full online parameter identification push the boundaries further: even when the identified parameters are far outside the defined uncertainty range, the controller maintains acceptable flight performance comparable to simulations, provided actuator time constants remain below 40 ms.

This work represents a significant step toward truly autonomous, throw-and-go quadrotor deployment. By closing the loop between real-time parameter estimation and robust control synthesis, the system can adapt to unknown payloads, environmental changes, or structural damage without manual tuning. The results were accepted to the ICUAS 2026 conference and promise practical benefits for search-and-rescue, package delivery, and other scenarios where drones must operate reliably in unpredictable conditions.