New stability analysis reveals the most dangerous model mismatch for VTOL UAV pitch control
Researchers derived a closed-form fifth-order transfer function to map stability regions for INDI controllers...
A new paper from ETH Zurich by Lorenzo Schenk, Guillaume Ducard, and Christopher Onder tackles a critical gap in flight control for tilt-rotor VTOL UAVs: how much model mismatch can an INDI (Incremental Nonlinear Dynamic Inversion) pitch-rate controller tolerate before going unstable? INDI is popular in drone control because it needs less aerodynamic modeling while still rejecting disturbances well, but its stability boundaries under parameter errors were not analytically pinned down—especially for the fast inner loop that handles pitch rate via elevons.
The team isolated the pitch-rate/elevon subchannel of a cascaded INDI controller and modeled the full controller–estimator–actuator–plant loop. They derived a closed-form fifth-order transfer function and applied the Routh–Hurwitz criterion across a parameterized linear model. Two representative three-parameter sweeps produced interpretable stability regions—feasibility maps that reveal exactly how much error in inertia, actuator lag, and control effectiveness the system can handle. The most striking result: a sign error in control allocation (e.g., commanding opposite elevon deflection) is the single most dangerous destabilizing factor, while moderate actuator delay and inertia mismatch cause little trouble when gains are nominal.
Based on these maps, the authors propose two uncertainty-aware tuning procedures. A robustness-oriented design maximizes a weighted worst-case combination of gain and phase margins, while a performance-oriented design maximizes worst-case closed-loop bandwidth subject to margin constraints. The work gives concrete recommendations: for conservative operations, prioritize avoiding control-effectiveness sign errors; for aggressive maneuvers, trade some bandwidth for phase margin. This is the first parameter-explicit, closed-form stability characterization for INDI on tilt-rotor VTOLs, offering a practical template for control engineers tuning similar architectures.
- Derived a closed-form fifth-order transfer function for the full INDI pitch-rate loop (controller+estimator+actuator+plant)
- Control-effectiveness mismatch (especially sign errors) identified as the most destabilizing factor; actuator lag and inertia errors are relatively benign at nominal gain
- Proposed two tuning procedures: robustness-oriented (maximizes worst-case gain/phase margins) and performance-oriented (maximizes bandwidth under margin constraints)
Why It Matters
Gives drone control engineers explicit stability boundaries and tuning guidelines for safer, more robust VTOL UAV flight control.