Autotuning the Unified Control and Generalized Control Allocation for VTOL Vehicles using DiffTune

Vertical takeoff and landing (VTOL) vehicles are gaining attention in advanced air mobility (AAM) because of their versatility and efficiency. Transitioning VTOL vehicles are designed to transition between different flight modes, where mode-specific control approaches are commonly used. Recently, a unified control design is proposed to utilize a single control structure across the entire flight envelope. While this approach has the advantage of control uniformity and simplicity in control implementation, the control parameters still need to be tuned jointly for a unified performance metric regarding the vehicle’s flight. Yet, the high dimensionality of the parameter space and the non-linearity of the cascaded control and allocation make tuning by hand challenging and demand more efficient tools to reduce the tuning effort and improve tuning quality than a trial-and-error type of manual tuning. DiffTune is a gradient-based autotuning method that can effectively tune high-dimensional nonlinear systems. In this study, we use DiffTune for the unified control design and generalized control allocation to automate the control tuning of VTOL vehicles. We leverage the Lift+Cruise vehicle in the Generic Urban Air Mobility (GUAM) simulator and demonstrate the efficiency and effectiveness of the proposed method compared with the state-of-the-art autotuning method. DiffTune achieves nearly minimal tracking error after the first iteration and reduces the tracking error by up to 87.5% compared to the average of the minimum error obtained by the baseline method.