The review describes nonlinear dynamic inversion (NDI) as a family of flight-control architectures with trade-offs rather than a single clear winner. In the PH-LAB flight-test example, incremental NDI had the best reported tracking but the highest actuator loads. Model-based NDI showed the opposite pattern, while hybrid NDI was a compromise between performance and actuator utilization.
The review is a preprint. It covers subsequent developments toward extended capabilities, as well as incremental, sensory and hybrid NDI forms from different perspectives. Alongside that review, the article describes a five-step process for deriving and implementing an NDI inverse-model core.
The trade-offs behind the designs
The authors conclude that NDI is better viewed as a modular architecture that can be combined with other control methods than as a stand-alone design method. The review compares the variants using inversion error, algorithmic complexity, disturbance compensation, control-failure handling and internal dynamics.
The comparison also considers how strongly each design depends on a vehicle model and how sensitive it is to measurements. Model-based NDI is characterized as highly model-dependent. Sensory or incremental NDI has lower model dependence but high sensitivity to output derivatives.
For incremental NDI, the review describes direct angular-acceleration and control-deflection feedback. It says that both performance and control activity depend on measurement quality.
Hybrid NDI blends higher-frequency model estimates with lower-frequency sensor measurements through a complementary filter. The review leaves the choice of how those two sources are combined to the application.
What the PH-LAB tests showed
The practical example uses PH-LAB, a Cessna Citation II CS-25-class research aircraft fitted with an experimental fly-by-wire system. The system used servos for the aileron, elevator and rudder.
The flight tests mainly used step responses in different axes and pilot-in-the-loop handling-quality maneuvers. The comparison examined roll-angle tracking and aileron utilization.
In the roll-angle plots, the reference and measured roll angles were essentially similar for all four variants. The overall comparison nevertheless reported that incremental NDI had the best tracking but the highest actuator loads, model-based NDI showed the opposite pattern, and hybrid NDI sat between them in performance and actuator utilization.
Differing conditions such as turbulence limited a more detailed comparison. Numerical values from the plots were not tabulated, and the review reports no numerical effect estimates or uncertainty intervals for the tracking-versus-load result.
For the described implementation, the base clock was 2000 Hz and most functions, including the flight controller, ran at 1000 Hz. Tests indicated that 100 Hz was sufficient for the developed control law. The hybrid controller used a first-order complementary filter with a time constant of 0.5 seconds.
A method used across many vehicle types
The review describes NDI applications in fixed-wing and tilt-wing aircraft, rotorcraft, drones and multirotors, eVTOL configurations, airships, and underwater or aerial/underwater robotic systems. Across those applications, the review continues to frame NDI as a set of architecture choices with different trade-offs.
The review’s conclusion is correspondingly architectural: NDI can serve as a modular element combined with other control methods. That framing leaves practical questions about how to balance tracking, actuator utilization and sensor reliance in different applications.
Paper data and sources
Original title: Dynamic Inversion: An Incrementally Evolving Methodology for Flight Control Design
Authors: Daniel Milz, Gertjan Looye
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-24
DOI: Not available
Original paper · Full text